THE USE OF VIRTUAL SELF MODELS TO PROMOTE SELF-EFFICACY AND
PHYSICAL ACTIVITY PERFORMANCE
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF COMMUNICATION
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Jesse Fox
May 2010
ACKNOWLEDGEMENTS
This dissertation is the culmination of four years of hard labor, but rather than an
ending, I like to consider it a beginning: I hope the ideas I have generated in this process
will lead to years upon years of fruitful research in my academic career. To that end, I
would like to thank a great number of people who have helped me throughout my
academic lifetime.
First, I would like to thank those who helped me sprout my intellectual roots. Mrs.
Mimi Southard, Mrs. Constance Hatfield, Mrs. Alice Brandon, and Ms. Joey Ansback
kept me from spending the entirety of my primary and secondary education staring out
the window in the back row of the classroom. I appreciate their endless patience with my
mental effervescence as I learned to love learning.
Next, I owe a great debt to those who kept me stimulated as an undergraduate at
the University of Kentucky. Thanks to Dr. Alan D. DeSantis, Ph.D., whose
Communication 101 class first lured me to the field and who I always hope to make
proud. Thanks to Dr. Ramona R. Rush, who changed my perspective of the world in onehour-fifteen-minute doses of media literacy and convinced me of the possibility of grad
school. I can say nothing that could possibly convey the role that Pat White played in my
college years as a devoted teacher and remarkable human being who forced me to
confront my writing and my life. May he rest in peace.
I never would have fallen in love with academics and committed myself to
professordom if it weren‘t for my time at the University of Arizona. Dr. Jake Harwood
taught me that numbers are good and was infinitely supportive of my teaching. Dr. Dale
Kunkel has continued to provide resources, insight, and encouragement. Dr. Tara
iv
Emmers-Sommer was responsible for piquing my interests with her thoughtful and
strenuous syllabi, and it was two of her fascinating classes on health communication and
sex and gender that led me to the research domains I study today. Dr. Dana Mastro has
continued to be both a mentor and a role model for me, and I thank her for all the sound
advice that has helped me survive my graduate career. Most of all, I must thank my
fellow UA Comm alumni, who made my two years in Tucson the best time a girl could
have outside of Kentucky, and continue to be the highlight of every conference: Tamar
Sarnoff, Carolyn Donnerstein, Priya Raman, Leah Kattau, Michael Coffman, Katie
Warber, Alesia Hanzal, Sam Dorros, Jeff Halford, and Michelle Ortiz, who I am glad to
be rejoining at Ohio State.
Here at Stanford, I would like to thank my Committee, Dr. Jeremy Bailenson, Dr.
Byron Reeves, Dr. Fred Turner, Dr. Vladlen Koltun, and my Chair, Dr. Abby King, for
taking the time to read through this dissertation and for being so supportive of this work.
I thank Dr. Reeves for being such a steady, positive, and supportive force during my time
here at Stanford and for teaching me the wonders of physiological research methods. I
would like to thank Dr. Turner for his absolutely fascinating courses and reading lists, for
keeping the English major inside from shriveling by cultivating my writing, and for
honing my qualitative research skills. I would like to thank my adviser, Dr. Bailenson, for
giving me the opportunity to conduct research at the Virtual Human Interaction Lab, one
of the most renowned research facilities in the field; for pushing me to reach for goals I
would have deemed unattainable; and for teaching me there is always a way to make
lemonade from lemons. I would also like to thank the Department of Communication
v
staff, especially Susie Ementon, for her guidance and willingness to assuage my
paperwork paranoia during this process.
This work would not have been possible without the contributions of other VHIL
programmers and students. Thanks to Chris Lin, Nick Yee, Joseph Binney, Tony
Ricciardi, Deonne Castaneda, Michelle Del Rosario, Steve Lesser, and Jade Wang for
their programming and technical assistance. Special thanks to Steven Duplinsky, whose
availability and assistance on the dissertation studies was invaluable. Thanks to my
research assistants over the years: Mandy Zibart, Stephanie Vezich, Alanna Nass, Todd
LaGuardia, Amanda Le, Claire Slattery, and Fiona Wainwright. Extra thanks to the long
hours from Katie Ambrose, Brittany Billmaier, and Liz Tricase, who were all rock stars.
Beyond the academic realm, I never would have made it here or much of
anywhere without my Kentucky folk. Thanks to Kiiiiim, Fur, and Sarah, the oldest of the
old, Kyle, and Dani, who is forever my girl in more than a Paula Abdul kind of way. It‘s
hard to believe that my friends from nerd camp have been around for half my life: Joe
Bun, Micajah, Brian, SetSet, Mot, David, Douglas Lee, and Brigid. Those who joined the
party at UK: Ann, Jamie Dawn, Mark, SKV, Bailey, Salyer x 2, Chrissy, Diane, Kirt,
Kristin, Amy, Diana, Brittain, and Renate. Y‘all are what make home home, and every
time I visit I feel rejuvenated on every level.
My California crew has helped make my time away from the homeland rather
sweet. Thanks to the rest of JAGH, Abhay Sukumaran, Sun Joo (Grace) Ahn, and Helen
Harris, for the constant support and empathy that only fellow graduate students can
provide. Thanks to Kathryn Segovia for her forever sunny view. My volleyball gang
quickly evolved from teammates into friends, including but not limited to Greg, Mark,
vi
Christian, George, Nadia, Jessica, Mary (and Mike), Karin, Ronda, Jenna, and Judy.
Thanks to the Silicon Valley Roller Girls for reigniting a spark I never knew had died.
Dr. Priya Raman deserves special acclaim for following me from Arizona to the South
Bay, keeping me well fed, and always being an intellectual sounding board.
Special thanks goes to my best friend, Joseph Bunny Villaveces, for sticking it out
and keeping me sane in Arizona and now California. And thanks to Amanda BommerVillaveces, and now Bebe Bommer-Villaveces, for keeping Joe sane.
Christian Zentgraf has been an eyewitness of this process for the last year, and I
must apologize for any sleeplessness it may have caused. He has provided me with the
bagels, coffee, and backrubs necessary to get through the 18-hour days at the computer.
Thanks for being there in the times when I need it most. I am forever grateful for his
support, companionship, and willingness to leave this great weather behind for Ohio.
Most of all, I would like to thank my parents, Darrell and Sherry Fox, who raised
me and my brother Denny to be real smart. As parents and teachers, you fostered my
academic growth from the get go. Thank you for always letting me check out as many
books from the library as I could keep tucked under my chin. Thanks for quizzing me on
my state capitals the night before my first kindergarten class. Thanks for letting me hang
out in your classrooms, grade your papers, clean out lockers, and mow the football field
in the summer so that I would love even the dirty work of school. Thank you for being
patient and tolerating every creative endeavor and wild misstep along the way. Thanks
for teaching me to be grateful and resilient and strong-willed. I was lucky enough to
inherit my intellect and a love of teaching from you, and so here I am…even though Dad
was probably right.
vii
TABLE OF CONTENTS
CHAPTER ONE: INTRODUCTION ..................................................................................1
CHAPTER TWO: SOCIAL COGNITIVE THEORY .........................................................4
CHAPTER THREE: TRADITIONAL MASS MEDIA HEALTH CAMPAIGNS ...........17
CHAPTER FOUR: HEALTH BEHAVIOR CHANGE USING NEW MEDIA
TECHNOLOGIES .............................................................................................................20
CHAPTER FIVE: VIRTUAL HUMANS AND VIRTUAL SELVES..............................24
CHAPTER SIX: VIRTUAL ENVIRONMENTS ..............................................................32
CHAPTER NINE: PRETESTS..........................................................................................48
Pretest 1: Vicarious Reinforcement of Virtual Selves ...................................................49
Pretest 2: Reward and Punishment of Virtual Selves and Virtual Others......................53
Pretest 3: Effects of Exercising Virtual Selves and Virtual Others After 24 Hours ......58
Pretest 4: Physiological Responses to Virtual Selves and Others ..................................65
Pretest 5: Presence and the Imitation of Eating Behaviors ............................................70
Justification for Dissertation Studies .............................................................................78
CHAPTER TEN: DISSERTATION STUDY ONE: DEGREES OF IDENTIFICATION
WITH VIRTUAL SELVES AND OTHERS .....................................................................80
Method ...........................................................................................................................82
Results ............................................................................................................................87
Discussion ......................................................................................................................96
CHAPTER ELEVEN: DISSERTATION STUDY TWO: COMPARING VIRTUAL
AND IMAGINED STIMULI FOR HIGH AND LOW SELF-EFFICACY
PARTICIPANTS .............................................................................................................100
viii
Method .........................................................................................................................106
Discussion ....................................................................................................................124
NOTES .............................................................................................................................140
APPENDIX A: STUDY 1 SURVEY ITEMS .................................................................142
APPENDIX B: STUDY 1 CONDITION x SEX FACTORIAL ANOVA RESULTS ....146
APPENDIX C: STUDY 1 SELF VS. OTHER (SIMILAR & DISSIMILAR) t-TEST
RESULTS ........................................................................................................................147
APPENDIX D: STUDY 1 SELF & SIMILAR VS. DISSIMILAR t-TEST RESULTS .148
APPENDIX E: STUDY 2 SURVEY ITEMS ..................................................................149
REFERENCES ................................................................................................................153
ix
LIST OF TABLES
Table 1. Factor Loadings for Physical Activity Items…………………………………...62
Table 2. Correlation Matrix for Factor Items…………………………………………….62
Table 3. Study 1 Means and Standard Deviations for Dependent Variables by Condition
and Sex…………………...………………………………………………………………89
Table 4. Pearson Correlations between Study 1 Dependent Variables (N = 75)………...95
Table 5. Study 2 Means and Standard Deviations for Dependent Variables by Condition
and Sex………………………………………………………………………………….114
x
LIST OF FIGURES
Figure 1. An immersive virtual environment setup……………………………………...33
Figure 2. Constructing the virtual self……………………………………………….…..46
Figure 3. Example stimuli for Pretest 1…………………………………………….……52
Figure 4. Results of Pretest 1………………………………………………………….....53
Figure 5. Results of Pretest 2…………………………………………………………….57
Figure 6. A running virtual human (A) and a loitering virtual human (B)………………59
Figure 7. Results of Pretest 3…………………………………………………………….64
Figure 8. The setup for physiological measures during immersion in the HMD……..…66
Figure 9. Results of Pretest 4…………………………………………………………….69
Figure 10. Example stimuli from Pretest 5………………………………………………73
Figure 11. Results of Pretest 5………………………………………………………...…77
Figure 12. Example stimuli for Study 1………………………………………...……….84
Figure 13. Specific self-efficacy by Condition and Sex…………………………………91
Figure 14. Effect of Condition on exercise repetitions after removal of unattractive
stimuli…………………………………………………………………………………....94
Figure 15. Example stimuli for Study 2……………………………………………...…110
Figure 16. Women‘s estimated pushups before and after treatment…………………....117
Figure 17. Women‘s perceived resemblance by Condition…………………………….119
Figure 18. Men‘s perceived resemblance by Condition………………………………..120
Figure 19. Women‘s identification by Condition………………………………………121
Figure 20. Men‘s identification by Condition………………………………………….122
Figure 21. Women‘s presence by Condition……………………………………………123
xi
CHAPTER ONE: INTRODUCTION
Research in the health domain provides strong support for the role of self-efficacy
in the adoption and maintenance of healthy behaviors. Self-efficacy is an individual‘s
perceived ability to enact a behavior and attain anticipated outcomes (Bandura, 1977,
1997). If a person feels that he or she can perform a behavior, he or she is more likely to
do it. For example, if a man feels it is impossible for him to lose weight, he is unlikely to
make an effort to do so; however, if he feels he is capable, he is more likely to start a diet
or exercise plan. Self-efficacy also helps an individual overcome perceived obstacles,
setbacks, and other negative events that may discourage behavioral modeling. For
example, if that same man loses some weight, but then regains a few pounds back, his
self-efficacy may help him recover and stick to his initial weight loss goals. Previous
studies have shown that self-efficacy plays a major role in smoking cessation, exercise
adherence, dieting, and other forms of disease prevention and management.
Self-efficacy is a driving force in social cognitive theory, which suggests that
people can learn behaviors vicariously through the observation of models (Bandura,
1977, 1986, 2001). Virtual environments (VEs) present the unique opportunity to create
and manipulate photorealistic virtual humans, which can be designed to look remarkably
similar to the self (Bailenson, Blascovich, & Guadagno, 2008). These virtual selves can
behave in ways that the self does not, enabling them to serve as models to promote
imitation of behaviors by the person in the physical world. My previous research has
supported the utility of other components of social cognitive theory in the development of
effective virtual models for diet and exercise. I have used virtual humans to demonstrate
that identification, the extent to which an individual relates to and feels similar to a
1
model, influences modeling behavior. I have manipulated virtual models to show
vicarious reinforcement, specifically the rewards (weight loss) and punishments (weight
gain) associated with exercise. My research has shown that virtual self-models, which
provide high levels of identification, lead to greater modeling of the portrayed behavior,
particularly those that demonstrate rewards and punishments. At this stage, further
research is needed to determine the mechanisms behind this modeling, particularly the
role of self-efficacy as it is considered such a fundamental aspect of health behavior
change.
In light of my earlier work in virtual reality, another lingering question is whether
these stimuli must be delivered in a virtual environment. Virtual environments can be
costly, creating virtual scenes can be time-consuming, and the equipment can be
cumbersome. Mental imagery, on the other hand, is a low cost and low investment
method. Perspectives on cognitive effort, including cognitive load theory (Chandler &
Sweller, 1991), indicate that these two stimuli may require different levels of cognitive
processing capacity. Engaging in mental imagery is deemed a more effortful and
cognitively demanding process than observing a visual stimulus, and may place
participants at a higher risk for cognitive overload. Additional research shows that some
people are not adept at visualizing; thus, mental imagery may not be a fruitful task for
them. Given the constraints of each type of stimulus, it is important to determine if there
is an advantage to either treatment in promoting health behaviors.
Because of the obesity epidemic and the continuous rise in death rates for heart
disease and preventable forms of cancer, many scientists have devoted their research
efforts to developing mass media campaigns and other mediated treatments to inform the
2
public and effect changes in health behavior. Traditional media campaigns are limited
however in their ability to deliver a specific, personalized, and impactful message to an
individual. Immersive virtual environments present the opportunity to create uniquely
tailored, interactive, rich, and engaging health messages. Considering the importance of
health on an individual and social level as well as the growing opportunities provided by
the diffusion of new media technologies, the affordances of VEs and their utility in the
delivery of health information merits further research. The studies of this dissertation
project extend this line of inquiry by considering the role of virtual and imagined selfmodels in changing individuals‘ health beliefs and behaviors. This dissertation will 1)
explain the role of self-efficacy within the context of social cognitive theory; 2) review
previous research on modeling health behaviors across media; 3) discuss virtual self
representations as well as theoretical frameworks that inform and describe their use; 4)
consider the attributes and affordances of immersive virtual environments; 5) describe the
pretests that inform the current studies; and 6) present two new studies designed to shed
more light on our understanding of virtual behavioral modeling.
3
CHAPTER TWO: SOCIAL COGNITIVE THEORY
Social cognitive theory (SCT), originally known as social learning theory, was
derived from behaviorism. Bandura (1977) challenged many of the primary assertions of
behaviorism, particularly the stimulus-response model of human behavior that suggested
―a one-way control process which reduces individuals to passive respondents to the
vagaries of whatever influences impinge upon them‖ (p. 6). Humans, Bandura argued, are
more complex creatures, capable of reasoning and individual creative response.
Regarding this human agency, Bandura (2001) noted that people are ―self-organizing,
proactive, self-reflecting, and self-regulating, not just reactive organisms shaped and
shepherded by environmental events or inner forces‖ (p. 266). Thus, social learning
involves active, agentic participation in cognitive decisions and the behaviors that result
(Bandura, 1989).
Because of this conceptualization, Bandura (1977) also rejected the one-way flow
of stimulus-response models. Instead, he posited that personal determinants, behavioral
determinants, and environmental determinants interacted and exerted mutual influence
over each other, and that this triadic reciprocal determinism served as a better predictor
of human behavior (Bandura, 1986). These factors are not necessarily balanced, nor do
they necessarily occur at the same time. In a given instance, any one may serve as the
dominant factor and override the effects of the others. Regardless, it is these three factors
that determine the viability of social learning taking place.
According to SCT, the observation of modeling is a learning process from which
the observer gleans information about the modeled behavior. That information is
processed in a symbolic manner and then stored for future use. Four processes are needed
4
for observational learning to occur. Attention is the ability of the individual to perceive
the behavior, including the individual‘s sensory capacities and perceptual sets; relevant
traits of the observed behavior such as functional value, salience, and complexity; and the
attractiveness and distinctiveness of the model (Bandura, 1977). Retention, or the ability
to create mental symbolic representations for future recall, involves remembering both
verbal coding and mental images pertaining to the behavior. This allows the individual to
cognitively organize the information for future access as well as mentally rehearse the
behavior. Motor reproduction, on the other hand, involves the physical enactment of
observed behaviors. Finally, motivation is essential; through external, vicarious, and selfreinforcement, the individual determines whether to enact or avoid the learned behavior
(Bandura, 1977). These four processes determine what symbolic representations will
exist regarding the referent behavior and what information will guide decision-making
regarding enactment.
The subsequent imitation of these learned behaviors is contingent on four factors.
Identification refers to the extent to which an individual relates to a model and feels that
s/he is similar to the model (Bandura, 1977; Bandura & Huston, 1961). Vicarious
reinforcement suggests that individuals need not experience rewards or punishments
themselves in order to learn behaviors; rather, they can observe and interpret the
consequences experienced by a model and make inferences as to the likelihood of
incurring these outcomes themselves (Bandura, 1977; Bandura, Ross, & Ross, 1963).
Observing this reinforcement helps the observer develop outcome expectancies, or beliefs
in the likelihood of the behavior yielding particular outcomes (Bandura, 1977). Another
essential component of SCT is self-efficacy, the individual‘s perceived ability to enact a
5
behavior and attain anticipated outcomes (Bandura, 1977, 1982, 1997). The belief that
one is capable of fulfilling one‘s goals (i.e., successful imitation) is crucial in determining
whether or not one will attempt to mimic the behavior. Each of these components will be
reviewed individually.
Identification
Identification refers to the extent to which an individual relates to a model and
feels that s/he is similar to the model. Kelman (1961) originally recognized identification
as one of three processes of social influence. Identification has been shown to increase
the likelihood of performing learned behaviors (Bandura & Huston, 1961; Bandura,
2001; Schunk, 1987). Observers must feel that the model is similar enough to them that
they are able to experience the same outcomes. Similarity may be based on physical
traits, personality variables, or shared beliefs and attitudes (Stotland, 1969). Indeed, the
likelihood of learning increases when models are of the same sex (Andsager, Bemker,
Choi, & Torwel, 2006; Bussey & Perry, 1982; Kazdin, 1974), age (Kazdin, 1974; Schunk
& Hanson, 1985), race/ethnicity (Anderson & McMillion, 1995; Kalichman, Kelly,
Hunter, Murphy, & Tyler, 1993) or skill level (Meichenbaum, 1971), as well as when
models demonstrate similar opinions (Hilmert, Kulik, & Christenfeld, 2006),
backgrounds (Rosekrans, 1967), or previous behaviors (Andsager et al., 2006).
Degree of identification with a model has also been shown as an important factor
in ascertaining the differential effects of media messages on individuals (Andsager,
Austin, & Pinkleton, 2001; Basil, 1996; Cohen, 2001, 2002; Maccoby & Wilson, 1957).
Many mediated messages about health behaviors use models similar to the target
audience to foster identification. Entertainment-education efforts typically match the
6
characters serving as models with the target audience in terms of physical characteristics
as well as beliefs, attitudes, and behaviors in order to achieve the highest levels of
identification, which in turn are expected to impact modeling behavior (Singhal &
Rogers, 1999, 2002). Ito, Kalyanaraman, Brown, and Miller (2008) created an interactive
CD-ROM that offered its female adolescent participants their choice of avatars to guide
them through information about sexually transmitted infections; over 60% of participants
chose avatars of the same race/ethnicity. It is likely the participants identified more
highly with these guides, and the authors speculated that this may influence message
effectiveness. Models matched on sex and ethnicity were more effective than nonmatched
models in getting African-American women to get tested for HIV and request condoms
(Kalichman et al., 1993); a subsequent study also revealed that African-American women
who viewed a model matched on sex and ethnicity was more effective than a model
matched solely on ethnicity in persuading them to get tested (Kalichman & Coley, 1995).
A recent media campaign to encourage healthy eating and physical activity among New
Orleans residents featured African-American actors speaking with a local accent in order
to promote higher levels of identification among the target audience (Beaudoin,
Fernandez, Wall, & Farley, 2007). Andsager et al. (2006) found that perceived similarity
to a model in an anti-alcohol advertisement was positively related to the message‘s
effectiveness.
Although these mass media messages can target narrower groups, they cannot
achieve highly specific models for each individual audience member. Thus, it may be that
some of the message‘s potency is lost by the individual‘s inability to identify with a given
7
model. New technologies that will be discussed in the following chapter present
researchers with the opportunity to maximize identification with a model.
Vicarious Reinforcement
Bandura critiqued the behaviorist claim that learning necessitated direct
experience. For example, if a child touched a hot stove, she would be burned and thus
learn that a hot stove should be avoided in the future. Social cognitive theory argues that
observational learning could also occur: a stimulus could be encountered by an observed
other and yield a response. If the child observed another person touch a hot stove and get
burned, she would learn to avoid the behavior without having to experience it firsthand.
In addition, distal models of these consequences can be presented via media. This process
of vicarious reinforcement is a significant advancement offered by SCT. Individuals need
not experience rewards or punishments themselves in order to learn behaviors; rather they
can observe and interpret the consequences experienced by a model and make inferences
as to the likelihood of incurring these outcomes themselves (Bandura et al., 1963;
Bandura, 1965). For example, Bandura et al. (1963) noted that children who observed a
model rewarded for aggressive behavior were much more likely to imitate that behavior
than children who observed a model punished for the same behavior. Bandura (1977)
clarifies that such reinforcement is not necessarily an automatic stimulus; rather,
reinforcement ―serves principally as an informative and motivational operation rather
than as a mechanical response strengthener‖ (p. 21). Vicarious reinforcement is effective
because it relies on the ability of observers to role play by envisioning a model‘s
behaviors—and the consequences thereof—as their own.
8
Outcome Expectancies
Based on this process, an individual develops outcome expectancies, or beliefs in
the likelihood of a certain behavior yielding particular outcomes (Bandura, 1977). If
modeled behavior consistently results in rewards, then observers will expect the outcome
of their own imitative behaviors to yield rewards. These rewards may create a
disinhibitory effect that then encourages the behavior regardless of other social
constraints. For example, although aggression is generally a socially unacceptable
behavior, if an observer sees that it is rewarded, disinhibition will occur and he or she
will likely imitate it. Conversely, if a model is punished for a behavior, observers will
anticipate that if they enact the behavior, they too will be punished. This typically
produces an inhibitory effect, in which the threat of punishment discourages a behavior.
Bandura (1977) found that if a behavior has anticipated consequences, a lack of reward
will serve as positive reinforcement if punishment is expected and as punishment if a
reward is expected. Hence, if a child breaks a known rule and is not punished, this
nonreward will still serve to positively reinforce the rule-breaking behavior. If the child
performs a good deed in anticipation of a reward and is not rewarded, the child will react
as if this were a punishment and be deterred from performing the good deed in the future.
Self-Efficacy
Bandura (1977, 2006) claimed that humans are agentic and self-reflective beings,
and thus the likelihood of behavior imitation is contingent on self-perception. Selfefficacy is an individual‘s perceived ability to enact a behavior and to attain anticipated
outcomes (Bandura, 1977). If an individual does not feel capable of imitating a modeled
behavior or achieving the observable outcome, he or she is less likely to attempt the
9
behavior. Self-efficacy beliefs are distinct from outcome expectancies; self-efficacy is
comprised of beliefs regarding the individual and the behavior, whereas outcome
expectancies delineate beliefs regarding the behavior and the outcome without
consideration of the individual. To clarify, an individual may be motivated to imitate a
behavior after observing the model‘s reward for this performance. The individual has an
expectation that the behavior will yield a certain outcome. However, a motivated
individual must also assess his or her own likelihood to be able to achieve that outcome;
thus, one‘s self-efficacy beliefs determine whether or not the outcome expectancy holds
and, as a result, whether or not the modeled behavior is imitated (Maibach & Murphy,
1995). Additionally, self-efficacy can both affect behavior and be affected by behavior.
For example, if a volleyball player feels he is able to perform a jump serve and
successfully jump serves on several occasions, his success in that behavior will enhance
his feelings of self-efficacy and increase the likelihood that he will jump serve in the
future.
Although there are many misconceptions in the literature, it is important to note
that self-efficacy is not a universal, unwavering belief or personality trait (Bandura, 1997,
2007). Rather, self-efficacy beliefs are specific to a particular behavior and can vary
according to contextual factors. As Bandura and Schunk (1981) stated, ―judgments of
self-efficacy are not simply reflectors of past performance…rather they reflect an
inferential process in which the self-ability inferences drawn from one‘s performances
vary, depending on how much weight is placed on personal and situational factors that
can affect how well one performs‖ (p. 596). Thus, measurement of self-efficacy must be
specifically cultivated to the context and behavior.
10
Operationalization and Measurement of Self-Efficacy
Because of misperceptions about the construct, many measures of self-efficacy
are too broad or incomplete to be useful to research in the tradition of social cognitive
theory (Bandura, 2007; Maibach & Murphy, 1995). Bandura (1986) suggested that
effective measurement of the construct should take into consideration level, strength, and
generality. The level of self-efficacy is determined by whether or not someone feels that
s/he can perform a given action, and if so, at what degree of difficulty. Perceived selfefficacy is also measured on the strength of the belief, or how certain the individual is
that s/he can perform the behavior. Finally, self-efficacy beliefs are evaluated on their
generality, or the scope of application either within a single behavioral domain or across
domains. Bandura (2004) also noted that a measurement instrument should include
questions that specifically address the impediments that may hinder an individual‘s selfefficacy (e.g., a mood state or an environmental issue like bad weather). Considering all
these facets ensures the utility and validity of the measurement and enables comparisons
to other relevant research.
Factors Determining Self-Efficacy
Bandura (1977, 1997) identified several mechanisms that may influence perceived
self-efficacy. Performance accomplishments or mastery experience are based on one‘s
own past efforts; if an individual has previously succeeded at a related task, self-efficacy
may be higher than if failure occurred. Thus, one‘s history with a specific behavior as
well as with related behaviors may come into play. For example, if a former tennis player
decides she wants to take the sport up again, she may recall her previous successes and
feel she is physically capable of returning to the court. Alternatively, an inexperienced
11
athlete may decide he wants to take up tennis, but is hindered by the fact that the last time
he tried to take up a new sport, he did not succeed. Physiological and psychological
feedback, often affective in nature, may also impact self-efficacy. High levels of arousal
may create a level of action readiness that promotes feelings of self-efficacy; in contrast,
one could have such a strong negative affective response, such as stress or anxiety, that
one feels less able to perform the behavior. (Bandura & Rosenthal, 1966; Thayer, 1989).
Based on the negative impact of such emotions on self-efficacy, Bandura (2001, 2002)
argues that fear appeals are not effective stimuli for promoting health behavior change
(for a contrary opinion, see Witte & Allen, 2000).
Persuasive efforts from others may also boost or detract from self-efficacy
(Bandura, 1977). A parent might offer encouragement to his child when she struggles
with her homework, promoting her sense of self-efficacy. Or, a basketball player might
deride and insult her opponent and convince him that he is incapable of shooting a free
throw. Vicarious experience can also enhance or diminish self-efficacy; by seeing others
perform the behavior successfully, individuals are more likely to assume they can do the
same. Situational circumstances, including variations in personal and environmental
factors, may play a role; for example, a vocalist might feel confident practicing with her
band in her garage, but if anyone drops in to listen, she might lose her confidence.
Bandura and Schunk (1981) found that proximal goal-setting, as opposed to making
distal goals or no goals, promoted students‘ self-efficacy and intrinsic interest in a subject
that previously did not interest them. Self-efficacy may also be affected by the level of
identification with an observed model (Bandura, 1997). If a heavyset woman cannot
identify with the size-2 model promoting the newest fad diet, she may feel that she cannot
12
possibly lose weight on that diet. In contrast, if the model looks very similar to the self, it
is possible that this may promote greater feelings of self-efficacy. Peng (2008) found that
greater identification with a video game character in a health-promoting video game led
to greater self-efficacy regarding dietary behaviors.
Outcomes of Self-Efficacy
Bandura (1986) identified several consequences of perceived self-efficacy. First,
it creates thought patterns or beliefs. The individual sets goals in accordance to perceived
ability, and people with a higher sense of self-efficacy are more likely to set higher goals
for themselves (Bandura, 1989). These higher goals may become a self-fulfilling
prophesy and result in greater achievement. Second, self-efficacy beliefs have notable
emotional effects. People with low self-efficacy are susceptible to stress and depression
(Bandura, 1986). Third, self-efficacy directly impacts an individual‘s effort. People with
a greater sense of self-efficacy are willing to expend more effort during goal attainment
(Bandura & Cervone, 1983). Finally, self-efficacy promotes persistence. People with
higher self-efficacy are more likely to stay motivated and maintain behaviors, even in the
face of adversity. They also recover more quickly from setbacks (Bandura, 1989). These
factors all play a role in health behavior maintenance via self-regulation.
Health Self-Efficacy
In the realm of health, self-efficacy is often cited as one of the primary causal
variables in the study of health behaviors. As Bandura (2000) stated, ―The stronger the
instilled perceived self-efficacy, the more likely are people to enlist and sustain the effort
needed to adopt and maintain health promoting behavior‖ (p. 304). Self-efficacy has been
shown to play a role in physical activity adoption and maintenance (Anderson, Wojcik,
13
Winett, & Williams, 2006; Garcia & King, 1991; Sallis et al., 1989; Sallis, Haskell,
Fortmann, Vranizan, Taylor, & Solomon, 1986; Rovniak, Anderson, Winett, & Stephens,
2002), diet (Anderson, Winett, & Wojcik, 2007; Schwarzer & Renner, 2000), physical
rehabilitation (Schwarzer, Luszczynska, Ziegelmann, Scholz, & Lippke, 2008); weight
loss (Linde, Rothman, Baldwin, & Jeffery, 2006; Warziski, Sereika, Styn, Music, &
Burke, 2008), contraception use (Forsyth & Carey, 1998; Wulfert & Wan, 1995), sun
safety (Jackson & Aiken, 2006; Myers & Horswill, 2006), breast self-examinations
(Garcia & Mann, 2003; Van Ryn, Lytle, & Kirscht, 1996), health information seeking
(Kennedy, O‘Leary, Beck, Pollard, & Simpson, 2004; Rains, 2008; Rimal, 2001a,
2001b), blood donation (Armitage & Conner, 2001), dental flossing (Schwarzer, Schüz,
Ziegelmann, Lippke, Luszczynska, & Scholz, 2007), STD prevention (O‘Leary, Jemmott,
& Jemmott, 2008), AIDS prevention (Maibach & Flora, 1993), smoking cessation
(Gwaltney, Metrik, Kahler, & Shiffman, 2009), drug and alcohol rehabilitation (Hyde,
Hankins, Deale, & Marteau, 2008), health behaviors during pregnancy (Moore, Turner,
Park, & Adler, 1996), seat belt use (Schwarzer et al., 2007), and disease management
(Clark & Dodge, 1999).
Some consistent effects have been shown for the role of self-efficacy in behavior
change; however, these are often limited to a certain domain (e.g., exercise or smoking)
and other models may be appropriate for other topics (Clark & Dodge, 1999; Van Ryn et
al., 1996). Self-efficacy has been shown in many studies to have direct effects on
behavior and also to serve a mediating role between other variables or types of treatment
and behavior outcomes. Rimal (2000) found that self-efficacy mediates the relationship
between knowledge about healthy behaviors and the performance of those behaviors.
14
Self-efficacy often mediates the relationship between the individual‘s outcome
expectancies and performance of the behavior (Schwarzer et al., 2007). In tests of the
theory of planned behavior, self-efficacy has been shown to affect behavioral intentions
and action (Rimal, 2001b; Van Ryn et al., 1996), sometimes playing a mediating role
between the two (Schwarzer et al., 2008). People with higher levels of self-efficacy are
more likely to enact health behavior changes than those with lower self-efficacy
(Bandura, 1998; Strecher, DeVellis, Becker, & Rosenstock, 1986). Those with high selfefficacy are also more likely to maintain exercise behaviors after their adoption (Sallis et
al., 1986; Garcia & King, 1991). Rimal (2001a) identified a mutual reinforcement model:
prior self-efficacy and knowledge predicted subsequent exercise behavior, and prior
exercise behavior predicted subsequent self-efficacy and knowledge.
Within the health context, researchers have attempted to integrate existing health
behavior stage-of-change theories with the construct of self-efficacy to identify temporal
and event-based fluctuations (Marlatt, Baer, & Quigley, 1995; Wallace, Buckworth,
Kirby, & Sherman, 2000). For example, Schwarzer (1992, 1999) proposed the health
action process approach (HAPA), a model of self-regulation that distinguishes between
an initial motivation phase, in which the intention to act is developed, and the volition
phase, in which actions, persistence, and possible failure and recovery take place. Based
on these two phases, Schwarzer distinguished between action self-efficacy (beliefs that
exist in the motivation phase, before behavior is enacted) and coping or recovery selfefficacy (beliefs that exist once the behavior has transpired, and maintenance is desired).
Schwarzer and Renner (2000) found that action self-efficacy predicted intentions during a
pretest assessment, and coping self-efficacy mediated the relationship between outcome
15
expectancies and behavior at posttest. Identifying these nuances in the construct may help
explain some of the disparate findings regarding the role of self-efficacy in health
behavior change.
16
CHAPTER THREE: TRADITIONAL MASS MEDIA HEALTH CAMPAIGNS
Vicarious reinforcement has been used to demonstrate the benefits and risks
associated with health-related behaviors. For example, showing negative consequences in
public health campaigns is expected to discourage observers from smoking or drug abuse
by showing models punished with appalling physical symptoms or harmed social
relationships (Witte & Allen, 2000). Entertainment-education efforts rooted in the
principles of SCT often portray the rewards and punishments associated with health
behaviors through plot lines in television and radio shows (Brown & Cody, 1991; Papa et
al., 2000; Rogers, Vaughan, Swalehe, Rao, Svenkerud, & Sood, 1999; Singhal & Rogers,
1999, 2002). For example, a recent story arc on the television drama ER addressed
adolescent obesity by featuring a doctor advising an obese teen with high blood pressure
to eat more fruits and vegetables and get more exercise (Valente, Murphy, Huang, Gusek,
Greene, & Beck, 2007). Vicarious reinforcement of these mediated models led to positive
changes: compared to non-viewers, viewers reported exercising more, eating more fruits
and vegetables, and were more likely to get their blood pressure checked.
Meyerowitz and Chaiken (1987) considered four ways that health messages could
affect health habits, including fear arousal, promotion of perceived vulnerability,
information transmission, and promotion of self-efficacy. They discovered that health
messages were effective in changing behaviors when they boosted the individual‘s sense
of self-efficacy. A well-developed media campaign should thus consider how its content
and delivery can best promote self-efficacy and self-regulation so that health behavior
change can endure.
17
Despite considerable theoretical speculation in the literature, mass media health
campaigns are not always rooted within solid theoretical predictions; instead, they are
developed like product advertising campaigns, relying on aesthetics, heuristics, and other
tricks of the trade that are effective for branding but not necessarily successful health
interventions. This may explain why so many of them fail to change audience behaviors
(Fishbein & Cappella, 2006; Flora & Thoresen, 1988). More recently, these efforts have
been entrusted to communication and social psychology scholars to ensure that efforts are
grounded in persuasive theories. Implementing the appropriate variables, several
campaigns have reported significant changes in health-related cognitions, attitudes, and
behaviors. The Stanford Five-City Multifactor Risk Reduction Project entailed a
multiyear media intervention with several assessments comparing control and treatment
cities on various health behaviors (Farquhar et al., 1985, 1990; Rimal, Flora, & Schooler,
1999; Schooler, Chaffee, Flora, & Roser, 1998; Schooler, Flora, & Farquhar, 1993;
Schooler, Sundar, & Flora, 1996). Several forms of media were implemented, including
printed tip sheets and booklets, a newspaper column, television shows, and televised
public service announcements. Findings included that once awareness was piqued via
media exposure, more health information-rich media were sought (Schooler et al., 1993).
Also, media exposure led to interpersonal discussion and information seeking. The
resultant knowledge and social support, in turn, promoted self-efficacy and behavioral
change (Rimal et al., 1999; Schooler et al., 1993).
Other mass media health campaigns have also achieved some success. People
exposed to the BBC‘s ―Fighting Fat, Fighting Fit‖ campaign reported significant weight
loss, less fat intake, and less snacking, as well as increases in exercise and fruit and
18
vegetable consumption over the course of the campaign (Miles, Rapoport, Wardle,
Afuape, & Duman, 2001). Renger, Steinfelt, and Lazarus (2002) targeted a population
using local participants and scenery in a PSA campaign to promote physical activity.
They found that the media campaign caused changes in viewers‘ perceptions of benefits
and barriers to physical activity as well as bolstering their self-efficacy and yielding an
increase in exercise behavior. Beaudoin et al. (2007) adopted similar targeting tactics to
accommodate an African-American population in New Orleans, and over five months the
campaign succeeded in improving attitudes toward a healthy diet and promoting more
walking behavior. In contrast to the success of these campaigns, Marcus, Owen, Forsyth,
Cavill, and Fridinger (1998) reviewed 28 media-based physical activity interventions and
found that they had very little impact on physical activity. Thus, although some
campaigns stand out as successful, it seems that in the grand scope, only a fraction of
them are achieving the desired effects.
19
CHAPTER FOUR: HEALTH BEHAVIOR CHANGE USING NEW MEDIA
TECHNOLOGIES
New technologies have the potential for many advances over traditional mass
media campaigns. Schooler et al. (1998) evaluated traditional health communication
campaign media on their key factors: reach, or how broad an audience they could
achieve; specificity, or how detailed and individualized the information could be; and
impact, or how effective the message was in yielding actual behavioral change. They
found that tip sheets, booklets, television programs, newspapers, and television public
service announcements all suffered from ―tradeoffs‖—that is, no one channel was able to
reach a great number of people, present a specific, targeted message, and successfully
change health behaviors.
Evolving computer-based media have the potential to overcome these limitations.
The use of the Internet and computer-generated interventions is often less expensive,
faster in both the creation and delivery of messages, and able to reach larger numbers of
people (Napolitano & Marcus, 2002). Rather than pushing a generic message to the
broadest possible audience, messages can be effectively and efficiently targeted and
tailored to cater to the individual‘s needs (Kreuter & Strecher, 1996; Kreuter, Strecher, &
Glassman, 1999). King, Ahn, Atienza, and Kraemer (2008) define targeting as ―the
systematic matching of subpopulations of individuals to specific interventions based on
the characteristics of the subpopulation,‖ whereas tailoring is ―adapting a particular
intervention that individuals are receiving to the specific needs of the individual‖ (p.
252). A campaign may be targeted towards a certain group (e.g., teenagers) and
distributed uniformly (e.g., through an advertisement that is designed specifically for
20
adolescents). Tailoring would require more specific information about an individual‘s
relevant health behaviors or his or her reactions to a particular intervention. Traditional
media efforts have been successful at targeting populations, but the interactivity of new
media is almost necessary to effectively tailor a mass health intervention.
Computer-Based Message Tailoring
Despite some successes, traditional mass media campaigns are hindered in their
ability to create appropriate stimuli for the individual. A recent focus of public health
interventions has been on the notion of tailoring mass media messages to create
individual-specific treatments (Bock, Marcus, Pinto, & Forsyth, 2001; Hawkins, Kreuter,
Resnicow, Fishbein, & Dijkstra, 2008; Napolitano & Marcus, 2002; Noar, Benac, &
Harris, 2007; Winett, Tate, Anderson, Wojcik, & Winett, 2005). Kreuter and Strecher
(1996) delivered tailored messages regarding health risks and found that tailored
messages were 18% more effective than nontailored communication in getting recipients
to change a risky health behavior. Campbell, DeVellis, Strecher, Ammerman, DeVellis,
and Sandler (1994) administered computer-generated tailored dietary information,
nontailored dietary information, or no information to a sample of adults. Those who
received the tailored intervention decreased their total fat by 23%, whereas the
nontailored group decreased intake by 9% and the control group by 3%. Marcus, Bock,
Pinto, Forsyth, Roberts, and Traficante (1998) and Bock et al. (2001) used individual
assessments of motivational readiness, self-efficacy, and current health behaviors to craft
tailored messages about increasing physical activity. After six months, those who
received tailored messages were engaging in significantly more exercise than those who
received a generic intervention.
21
Bandura (2004) suggests that interactive tailoring is the greatest strength of new
media. Computer-based feedback can be specified to accommodate the individual‘s
motivations, current self-efficacy levels, unique obstacles and setbacks, and progress
towards goals. DeBusk et al. (1994) developed the self-management model, enabling
users to monitor and self-regulate their health behaviors through the use of an interactive
computer system. First, goals, risk factors, obstacles, and motivations are identified and
integrated into the individual‘s recommended treatment plan. As they follow the tailored
advice, their progress is tracked, and users receive regular feedback on the effectiveness
of their actions. This continuous monitoring and feedback encourages self-regulatory
behaviors that support the desired health behavior change (DeBusk et al., 1994).
Indeed, because of the possibilities of interactivity and engagement with
computer-based messages (Bucy & Tao, 2007; Sundar, 2007), it is likely that they also
have the potential for greater impact (Goran & Reynolds, 2005). Interactive systems may
also promote greater adherence to long-term interventions as the opportunity for feedback
keeps the user invested and involved (Baranowski et al., 2003). With these capabilities, it
appears that Internet-based applications may resolve many issues associated with
traditional media.
Mobile Technologies
Extensive research has been conducted regarding the use of computer-generated,
phone-delivered health interventions to promote healthful behaviors (e.g., King et al.,
2007). Recent research has expanded this notion by using mobile devices to communicate
health messages. King, Ahn, Oliveira, Atienza, Castro, and Gardner (2008) provided a
group of over-50 adults with personal data assistants (PDAs) that helped monitor
22
participants‘ physical activity levels over eight weeks. Following Bandura, selfregulation was promoted through the provision of individualized feedback every day and
weekly goal-setting. Despite the sample‘s relative inexperience with PDAs, the devices
were effective at promoting exercise: those who received the intervention reported
greater caloric expenditure and more time engaging in moderate or vigorous physical
activity than the control group. A related study used a similar design and sample to
ascertain the effectiveness of PDAs for enhancing dietary choices (Atienza, King,
Oliviera, Ahn, & Gardner, 2008). Participants who were equipped with PDAs showed an
increase in vegetable intake over the treatment period, whereas those in the control group
did not. Future studies will undoubtedly continue to explore the utility of mobile devices
in delivering health messages, perhaps incorporating some of the content that will be
discussed next.
23
CHAPTER FIVE: VIRTUAL HUMANS AND VIRTUAL SELVES
Representations of interactants in virtual environments can vary from a high
fidelity virtual human to an anthropomorphized animal in an online role-playing game
(see Nowak & Rauh, 2006, for a review), and this representation can have effects on both
the user and observers (Castronova, 2004, 2005; Yee & Bailenson, 2007; Yee, Bailenson,
& Ducheneaut, 2009). When virtual representations resemble the human form, users
often treat them similarly to how they treat real people (Bailenson, Blascovich, Beall, &
Loomis, 2003; Garau, Slater, Pertaub, & Razzaque, 2005). Beyond their appearance,
these representations are distinguished by who or what controls their actions. Avatars are
controlled by a human user, whereas agents are controlled by an algorithm (Bailenson &
Blascovich, 2004). When a virtual representation is controlled by an algorithm, it is
referred to as an embodied agent (Cassell, 2000). Users distinguish between
representations that are controlled by humans as opposed to those controlled by agents; in
general, if users believe they are interacting with an avatar rather than an agent, they are
more persuaded and learn more (Hoyt, Blascovich, & Swinth, 2003; Okita, Bailenson, &
Schwartz, 2008).
Recent scholarship has focused on the use of virtual humans as persuasive agents.
Gilliath, McCall, Shaver, and Blascovich (2008) found that people experience feelings of
identification with and empathy toward virtual humans which may increase their
effectiveness as models and persuasive agents. Guadagno, Blascovich, Bailenson, and
McCall (2007) also noted that certain characteristics of the virtual human (in this study,
matching to the participant‘s sex) enhance its ability to persuade. Traditional social
facilitation effects have also been found with the use of virtual humans. Several scholars
24
have found that when people are in the presence of a virtual human, their performance on
easy tasks is enhanced, but performance of difficult tasks is hindered compared to being
alone (Hoyt et al., 2003; Park & Catrambone, 2007; Zanbaka, Ulinski, Goolkasian, &
Hodges, 2007).
Some studies have examined the role of persuasive agents specifically in healthrelated contexts. Ijsselsteijn, de Kort, Westerink, de Jager, and Bonants (2006) rendered a
coach within a virtual cycling environment and found that her presence led participants to
place a greater value on exercise and experience less tension and pressure, although she
had no effect on actual performance. Ito et al. (2008) found that participants relied on
self-similar avatars to guide them through information about sexually transmitted
infections. Skalski and Tamborini (2007) found that interactive agents were successful in
encouraging health message processing as well as promoting more positive attitudes
towards healthy blood pressure and reporting the intention to get one‘s blood pressure
checked. Bickmore, Gruber, and Picard (2005) found that interacting with a healthpromoting virtual agent over time led to more health-related information seeking. As
Bandura (1997) noted, persuasive messages from interpersonal sources can have a direct
impact on self-efficacy. Additionally, the self-efficacy effects of media messages are
often mediated by interpersonal means (Bandura, 2001). Using virtual humans to convey
these messages may maximize the impact of mediated messages because they evoke
many of the same feelings as interpersonal interactions.
Virtual Representations of the Self
Virtual representations of the self (VRSs) can also be powerful at evoking attitude
and behavioral changes. Avatars are often the method by which users can extend the self
25
into a virtual environment. Users are able to achieve goals such as social interaction, task
accomplishment, navigation, or game participation using their avatars. Because of the
flexibility of virtual environments, avatars allow users to experiment with their selfrepresentations, giving them the opportunity to selectively present features of their ―real‖
selves or experiment with different identities (Nakamura, 2002; Suler, 2002, 2004;
Taylor, 2002; Turkle, 1995; Walther, 1996, 2007; Yee, 2006). They can adopt a different
sex, gender, or sexuality; a different class or occupation; a different race or ethnicity; or a
different height, weight, or level of attractiveness. Users may be embodied in these
avatars for days, weeks, or even years in digital environments. Thus, it is unsurprising
that people report having strong feelings toward or experiences with their avatars (Lewis,
Weber, & Bowman, 2008).
Theoretical Explanations for the Effects of Virtual Selves
Relationships with virtual selves can be explored using a variety of frameworks
from the social psychological literature. Generally, these explanations consider 1) how
we choose to present ourselves and why; 2) how these presentations shape our sense of
self; and 3) how this identity shift alters our beliefs, attitudes, and behaviors both within
and outside of virtual environments.
Goffman‘s (1959) classic work integrated psychological and social
understandings of identity to consider how the individual interprets and presents oneself
in a social world. His dramaturgical notions of the front stage and back stage are
particularly relevant to how people choose to present themselves in virtual worlds.
Goffman claimed the front stage was where one performed one‘s identity, and the back
stage was where one could strip away the roles and performance and essentially be one‘s
26
―true‖ self. Because of the anonymity of online interactions, virtual environments present
the opportunity for those who are uncomfortable or unsatisfied with their front stage lives
to enact more aspects of their back stage selves.
Turkle (1995) suggests that the self is inherently fragmented, and that we can
explore the different parts of ourselves through alternative presentations and role play
online. Thus, beyond the expression of identity, virtual self representations may also
represent a new method of self-realization. By creating a virtual self, it may be possible
to become more attuned to one‘s real world self. In the postmodern view, the self is not a
unitary entity; rather, it is an amalgamation of multiple, fragmented selves. An avatar, or
multiple avatars, presents an opportunity to embody these fragmented selves in a process
of self-construction (Turkle, 1995).
This process of self-construction is not restricted to the ―true‖ real world self,
however. Markus and Nurius (1986) proposed the idea of possible selves, that is,
cognitive constructions of the self that are based on past experiences, future hopes, and
current goals and motivations. Our self-concept is not only what we are, but what we feel
we could possibly be. Virtual representations can be used to embody our wishful
identification selves (Hoffner, 1996; Hoffner & Buchanan, 2005), that is, the versions of
ourselves that we wish we could achieve (Suler, 2004). Indeed, Bessiére, Seay, and
Kiesler (2007) found that players‘ avatars in World of Warcraft more closely resembled
the idealized self than the real self. These wishful selves may also be projected into the
real world. Konijn, Bijvank, and Bushman (2007) found that wishful identification with a
video game character also carried over into real world interactions; the more adolescent
27
boys wished they were like the violent video game character they played, the more
aggressive behavior they demonstrated after game play.
How users are represented may impact identification, the bond users feel with
these representations, and the resultant effects on users‘ sense of self. Current
explorations of the concept of identification in the context of virtual worlds integrate the
work of Kelman (1961), Bandura (1977, 2001), and Cohen (2001). Klimmt, Hefner, and
Vorderer (2009) distinguish avatars (specifically those in video games) from other media
characters because of the interactivity and control that the user has over an avatar that is
not possible with a literary role or television character. Other conceptualizations of this
dynamic have described it as dyadic, maintaining that there is an inherent separation
between the observer and the observed character (e.g., transportation, Green & Brock,
2000; parasocial interaction, Horton & Wohl, 1956). Klimmt et al. argue that due to their
interactivity and the user‘s control over the character, video games create a monadic
relationship wherein ―players do not perceive the game (main) character as a social entity
distinct from themselves, but experience a merging of their own self and the game
protagonist‖ (p. 354). The authors thus define identification as ―a temporary alteration of
media users‘ self-concept through adoption of perceived characteristics of a media
person‖ (p. 356). They also argue that because of the active and responsive nature of this
form of identification, it is most similar in concept to role playing.
Furthermore, Klimmt et al. (2009) proposed that identification is selective and
temporally unstable. In virtual environments, characters may be far-reaching and
fantastical, pushing the boundaries of reality. Thus, users select specific traits with which
to identify. The consequences of these choices may explain differences in effects of video
28
games. For example, if a character conquers a battlefield and slaughters a hundred
enemies, a player might choose to identify with the character‘s courage, strength, power,
aggression, and/or ruthlessness. The traits the user adopts during the process of
identification may explain why some users experience increased feelings of aggression
after such game play whereas others do not (Anderson & Bushman, 2001; Anderson &
Dill, 2000; Anderson, Gentile, & Buckley, 2007). Indeed, some studies have found a link
between identification with violent game characters and aggression (Konijn et al., 2007)
and identification with characters leading to stereotyping and hostility (Eastin, Appiah, &
Cicchrillo, 2009). Identification is also noted to be temporally unstable. While playing a
game or immersed in a virtual world, the character might experience something that
diminishes the user‘s experience of identification. For example, a character might make a
choice the user disagrees with during a programmed narrative sequence, or the character
might lose a battle and appear weak. Thus, identification is a fluid process that can be
affected by many variables during an immersive experience and often impacts the effects
of these experiences (Bessière et al., 2007; Hefner, Klimmt, & Vorderer, 2007;
McDonald & Kim, 2001).
A related factor to the experience of identification is self-presence, or the
experience of feeling one‘s self within a virtual environment (Lee, 2004; Ratan, Santa
Cruz, & Vorderer, 2008; Ratan, 2010). Because one‘s physical body is not digitized and
thus limited in its ability to extend into and integrate with a virtual environment, it is
important to assess the degree to which people feel that they are able to interact as an
intact self within the VE. User controls, tracking systems, and avatars all serve as tools to
more naturally incorporate the self in the VE. Self-presence influences our experience of
29
immersion and may have a direct effect on outcomes within the virtual environment
(Ratan, 2010).
The degree to which we identify with or feel present within these avatars may also
reflect or impact the degree to which we incorporate pieces of them in our offline selves.
Self-perception theory (Bem, 1967, 1972) suggests that individuals infer their attitudes
and traits through the self-reflective observation of their own behaviors. For example, if a
woman spends a lot of time volunteering at a homeless shelter, she might infer that she is
a compassionate, sympathetic person who helps people in need. Goldstein and Cialdini
(2007) extended self-perception theory to incorporate modeled behaviors. They suggested
that individuals may observe the freely chosen behaviors of someone to whom they feel
close or have a ―merged identity‖ with and then infer their own attributes, a process they
labeled the spyglass self. For example, if the woman before observed her spouse
volunteering at the shelter, she may determine that since she and her spouse are so
similar, she is also charitably-minded, and this may increase her likelihood of also
volunteering at the shelter.
Self-perception theory relates to virtual selves because people may observe their
virtual self performing a behavior, and because they identify with and feel similar to the
virtual self, they may infer their own attributes based on that behavior and increase the
likelihood of them performing that behavior. Alternatively, they may observe the
behaviors of another virtual person who appears similar to them and with whom they feel
a merged identity, and the same effects may occur. An additional consideration should be
weighed, however: Goldstein and Cialdini also argue that the observer must believe that
the actions are a result of free choice. Due to the necessity of volition, it may be that if a
30
virtual representation‘s behaviors are clearly programmed and/or attributed to an agent,
the spyglass self effect does not occur, whereas if the behaviors are attributed to a humancontrolled avatar, the effect occurs.
Yee and Bailenson (2007) specifically applied self-perception theory within
virtual environments. The Proteus effect is hypothesized to occur when a user‘s virtual
self-representation is modified in a meaningful way that is often dissimilar to the physical
self. When the user then interacts with another person, the user‘s behavior conforms to
the modified self-representation regardless of the true physical self or the other‘s
impressions (Yee & Bailenson, 2007; Yee et al., 2009). For example, when participants
embody attractive avatars, they behave like attractive people, disclosing more personal
information and approaching another avatar more closely. When participants embody
taller avatars, they are more confident in a negotiation task (Yee & Bailenson, 2007).
Because virtual environments offer unlimited ways in which one can manipulate the
virtual self, the Proteus effect may explain many offline outcomes of online interactions.
In sum, these perspectives provide many explanations for the potential effects of
virtual self representations. Now, we will examine the different contexts in which we
encounter virtual selves.
31
CHAPTER SIX: VIRTUAL ENVIRONMENTS
Immersive virtual environments (IVEs), or what is generally understood as virtual
reality, enable researchers to investigate new mediated experiences and create novel
experimental simulations (Lanier, 1992, 2001). IVEs are defined by two characteristics:
the replacement of natural sensory information with digital information and the ability to
track and respond to users‘ movements in order to tailor that digital information (Loomis,
Blascovich, & Beall, 1999; Blascovich, 2001, 2002; Blascovich et al., 2002). One of the
most commonly implemented devices is a head-mounted display (HMD), a helmet or
headpiece with LCD screens fitted in front of the eyes that helps provide a wide,
stereoscopic view of the computer-generated environment. The image drawn inside the
HMD depends on the information given by the tracking apparatus. Various devices can
capture simple head movements, such as turning the head in different directions; the
position of the body in three-dimensional space (e.g., walking around a room); or body
movements, such as waving a hand or changing posture. Other sensory feedback may be
provided, such as sound incorporated via headphones or speakers, or touch using a haptic
device. Figure 1 illustrates the components of one type of IVE setup.
32
Figure 1. An immersive virtual environment setup. Cameras (A) in the corners of the room track
an optical sensor (B) as the participant moves around the room. An accelerometer (C) gathers
data on the participant‘s head movements. The data from these two tracking devices is sent to the
computer (D) which uses that information to render the room appropriately in the head-mounted
display (E).
IVEs have been used to replicate ―real world‖ studies, by simulating lifelike
environments and social interactions, as well as examine novel phenomena that are
enabled only in the virtual world (Fox, Arena, & Bailenson, 2009; Loomis et al., 1999).
IVEs have been used in a manner similar to traditional experimental environments to
study psychological processes such as interpersonal distance (Bailenson et al., 2003),
behavioral mimicry (Bailenson &Yee, 2005, 2007), leadership (Hoyt & Blascovich,
2007), social inhibition (Hoyt et al., 2003), obedience (Slater et al., 2006), prejudice
(Dotsch & Wigboldus, 2008) and persuasion (Guadagno et al., 2007). Other studies have
33
capitalized on the unique features of virtual environments to explore new affordances of
the technology, such as the ability to embody avatars with various physical traits (Groom,
Bailenson, & Nass, 2009; Yee & Bailenson, 2007; Yee et al., 2009); providing speakers
the ability to make eye contact with multiple interactants at the same time (Bailenson,
Beall, Loomis, Blascovich, & Turk, 2005); or optimizing the contextual factors of
learning environments by giving every student the same perspective in a virtual
classroom (Bailenson, Yee, Blascovich, Beall, Lundblad, & Jin, 2008). Here, a few
distinguishing features of VEs will be reviewed.
Media Richness
Media richness refers to the sensory quality of a medium and how it is
experienced by the user (Trevino, Lengel, & Daft, 1987). Daft, Lengel, and Trevino
(1987) compared media and face-to-face interactions on four criteria: 1) immediate
feedback; 2) transmission of multiple cues, such as nonverbal communication or
graphics; 3) language variety; and 4) personal focus. More richness is generally
associated with better outcomes (e.g., Cable & Yu, 2006; Scheck, Allmendinger, &
Hamann, 2008; Timmerman & Kruepke, 2006) although other factors can impact the
effectiveness of media rich environments (Rice, 1992). On the authors‘ range of high
(face-to-face interaction) to low (nonpersonalized text-based communication), IVEs
would be considered high on media richness, perhaps higher than any other form of
mediated communication. IVEs have the capacity for immediate feedback and
interactivity, as the user receives instant updating of the environment based on his or her
physical movements; also, real-time communication involving multiple cues is possible
via voice transmission, face-tracking to convey emotions, body movements tracked and
34
rendered on avatars, and other technologies. These methods also enable a variety of
language to be shared in IVEs as easily as it is in face-to-face communication. A virtual
environment can be continuously tailored to the individual based on the data it receives
(based on the participant‘s movements, physiology, or other information.) These
characteristics demonstrate that IVEs rank high on media richness based on Daft et al.‘s
criteria.
Presence
This high level of media richness is likely to influence how an individual
perceives and becomes engaged with a virtual environment. The experience of presence
entails the user‘s feelings that the mediated environment is real and that the user‘s
sensations and actions are responsive to the mediated world as opposed to the real,
physical one (Biocca, Harms, & Burgoon, 2003; Lee, 2004; Lombard & Ditton, 1997;
Loomis, 1992; Riva, Davide, & Ijsselsteijn, 2003; Slater & Steed, 2000; Steuer, 1992;
Wirth et al., 2007; Witmer & Singer, 1998). The user experiences presence as ―being
there‖ or ―losing oneself‖ in the mediated environment (Lombard & Ditton, 1997).
Although presence has been examined in the context of other media such as television
and books, because of the immersive nature of the virtual experience, it is of particular
importance to VE researchers. Presence may be a result of characteristics of the
technology used (Ijsselsteijn, de Ridder, Freeman, Avons, & Bouwhuis, 2001), aspects of
the environment such as graphic realism (Ivory & Kalyanaraman, 2007), or individual
differences among users (Sacau, Laarni, & Hartmann, 2008).
The examination of presence is important as previous studies have shown that the
subjective experience of presence can impact the effectiveness of virtual treatments
35
(Skalski & Tamborini, 2007; Villani, Riva, & Riva, 2007) and the degree to which these
stimuli translate into real world behavior (Fox, Bailenson, & Binney, 2009; Persky &
Blascovich, 2008; Price & Anderson, 2007). In a review of the research, Lee (2004)
identified three different aspects of presence, including physical, spatial, or
environmental presence (the feeling that you are in a particular virtual space; Lee, 2004),
social presence (the feeling that another person is sharing the virtual space with you;
Biocca et al., 2003), and self-presence (the experience of a virtual self-representation as
an extension of the self; Ratan et al., 2008; Ratan, 2010).
Interactivity
Participants inside an IVE often describe the experience as ―being in a movie.‖
Unlike a movie, however, an IVE is characterized by interactivity, which has been
characterized as an influential feature in new media (Bucy & Tao, 2007; Kiousis, 2002;
Leiner & Quiring, 2008; Liu & Shrum, 2003; Sundar, 2004, 2008; Vorderer, 2000).
Interactivity has been defined as both the user‘s perception that the medium is responding
to the user‘s actions (Leiner & Quiring, 2008; Steuer, 1992) as well as the medium‘s
technological capacity to actually respond (Lee, Park, & Jin, 2006). In an IVE, users have
the opportunity to engage with a responsive environment and virtual humans whose
behaviors are contingent on the user‘s actions. Thus, interactivity allows the user to be
both an observer and a participant in the environment, possibly leading to different or
more potent media effects (Vorderer, 2000). Early theories hypothesized that interactivity
would be a contributing factor to the experience of presence (Lombard & Ditton, 1997),
and subsequent empirical work has demonstrated this in the laboratory. Li, Daugherty,
and Biocca (2002) and Fortin and Dholakia (2005) found that participants exposed to
36
interactive advertisements reported greater presence than those exposed to noninteractive
ones. Skalski and Tamborini (2007) found that participants reported greater social
presence after experiencing an interactive agent rather than a noninteractive one. These
changes in the environment may help focus participants‘ attention and keep them more
engaged, resulting in greater presence.
37
CHAPTER SEVEN: VIRTUAL ENVIRONMENTS AND HEALTH
The ability to transform virtual environments makes them fruitful venues for
health interventions and research. Historically, virtual health treatments have evolved
from interactive CD-ROMs to specially designed video games, online virtual worlds, and
immersive virtual environments. The following section reviews the usage of VEs as a
tool for health behavior change.
Health-Oriented Video Games
Video games have been successfully implemented in the delivery of health
messages (Baranowski et al., 2003; Baranowski, Buday, Thompson, & Baranowski,
2008; Lieberman, 1997; Peng, 2009). Lieberman (1997) reported that asthmatic children
who played Bronkie the Bronchiasaurus, a game specifically designed to portray the selfmanagement of asthma, had significant gains in knowledge about asthma and their
perceived self-efficacy in managing their condition. Lieberman also found that
knowledge and self-efficacy persisted one month after children last played the game.
Baranowski and colleagues (2003) designed a video game to promote fruit and vegetable
consumption among children. After treatment, children who had played the game
reported consuming an additional serving of fruits and vegetables each day compared to
their non-game playing peers.
Peng (2008) argues that unlike other forms of media, video games offer players
the opportunity to become more engaged in the message via a mediated enactive
experience, or interactive role-playing. According to social cognitive theory, enactive
experience, in which the participant performs the behavior, is more effective in
generating self-efficacy than mere passive observation. Peng extends the concept of
38
enactive experience to entail the mediated version of playing a character in a game.
Indeed, she found that playing a health-promoting video game led to a greater increase in
self-efficacy than watching the game being played. Thus, involvement and interactivity
with virtual environments may lead to greater gains than other media formats.
Other Health-Oriented Virtual Environments
Virtual environments have long been employed in a variety of medical contexts,
particularly for training physicians. Virtual models of the human body have become
popular interactive tools for teaching medical students, nurses, and doctors the basics of
human anatomy as well as complicated surgical procedures (O‘Toole et al., 1998; Spitzer
& Ackerman, 2008; Thalmann & Thalmann, 1994). VEs have also been used to teach
medical personnel communication and decision-making skills because they can portray a
variety of situations, from a regular checkup to the chaos of an emergency room, that
practitioners may face (de Leo et al., 2003; Johnsen et al., 2006; Kenny, Rizzo, Parsons,
Gratch, & Swartout, 2007; Mantovani, Castelnuovo, Gaggioli, & Riva, 2003).
Because VEs proved to be useful in training medical personnel, soon they were
being developed as treatments for patients. One of the most common applications of fully
immersive virtual environments is virtual reality exposure therapy (VRET; Gregg &
Tarrier, 2007; Parsons & Rizzo, 2008; Powers & Emmelkamp, 2008; Riva, 2002, 2005;
Rothbaum, Hodges, & Kooper, 1997). Psychiatric researchers realized that IVEs
presented the opportunity to recreate highly realistic representations of fear-inducing
contexts to treat patients who suffer from a specific anxiety or phobias. In the virtual
environment, patients are gradually introduced to the negative stimulus in a virtual setting
until they become desensitized or are able to cope with their fear or anxiety. VRET has
39
been used to treat acrophobia (the fear of heights; Coelho, Santos, Silvério, & Silva,
2006), agoraphobia (fear of open spaces; Botella et al., 2007), arachnophobia (fear of
spiders; Cote & Bouchard, 2005); aviophobia (fear of flying; Rothbaum, Hodges, Smith,
Lee, & Price, 2000); public speaking anxiety (Harris, Kemmerling, & North, 2002), panic
disorder (Botella et al., 2007) and social phobia (Roy, Klinger, Legeron, Lauer, Chemin,
& Nugues, 2003). In addition to phobias, VRET has also been employed in the successful
treatment of combat-related post-traumatic stress disorder (PTSD; Reger & Gahm, 2008;
Rothbaum, Ruef, Litz, Han, & Hodges, 2003).
In the medical context, VEs have been shown to be an effective distraction
method for helping patients manage pain (Gold, Belmont, & Thomas, 2007; Hoffman,
Patterson, Seibel, Slotani, Jewett-Leahy, & Sharar, 2008). Children exposed to an
interactive distraction in an HMD as opposed to other forms of distraction significantly
increased their pain tolerance and pain thresholds (Dahlquist, McKenna, Jones, Dillinger,
Weiss, & Ackerman, 2007). These studies demonstrate the power of VEs and the
experience of immersion, as users who were present in a virtual environment were
actually able to block out unpleasantries of the physical environment.
VEs have also been explored as a tool for cognitive behavioral therapy.
Researchers have found that virtual cues can be used to stimulate alcohol cravings (Cho
et al., 2008) and nicotine cravings in cigarette smokers (Baumann & Sayette, 2006).
Thus, it is expected that these stimuli may be used therapeutically to teach addicts to cope
with craving-inducing cues in a variety of situations. VEs have also been used in studying
patients with eating disorders by exposing them to high-anxiety environments such as a
kitchen filled with fattening foods and examining patients‘ emotional reactions
40
(Gutiérrez-Maldonado, Ferrer-García, Caqueo-Urizar, & Letosa-Porta, 2006).
Researchers expect that these environments will be incorporated in therapy in which
patients learn to cope with anxiety-inducing situations in a healthy manner.
Physical Activity within Virtual Environments
Virtual reality has been used for over a decade as a tool for encouraging physical
activity and performance (Nigg, 2003). Many of these studies have focused on the use of
virtual environments to facilitate rehabilitation efforts (Schultheis & Rizzo, 2001;
Sveistrup et al., 2003). VEs have two features that uniquely facilitate physical
rehabilitation: the ability to capture and review one‘s physical behavior threedimensionally, thus enabling a close and interactive examination of one‘s progress and
failures, and the ability to see one‘s own avatar rendered in real time from a third-person
point of view (Bailenson, Patel, Nielsen, Bajscy, Jung, & Kurillo, 2008). VEs have been
used to help stroke victims regain a sense of balance while walking (Deutsch &
Mirelman, 2007; Fung, Richards, Malouin, McFadyen, & LaMontagne, 2006), help
children with cerebral palsy develop muscular coordination (Bryanton, Bossé, Brown,
McLean, McCormick, & Sveistrup, 2006), and study how virtual environments can help
the elderly restore coordination and balance after a fall, even in treacherous conditions
(Nyberg et al., 2006; Sondell et al., 2005).
Other studies have focused on using VEs to encourage longer and more vigorous
exercise sessions, thus increasing exertion and caloric burn. Annesi and Mazas (1997)
studied exercise adherence, comparing a VE-enhanced bicycle with a regular exercise
bicycle, and found that those exercising in the VE condition had a greater adherence.
Chuang and colleagues (2003) also compared cycling in a regular or virtual environment
41
and found that participants in the virtual environment cycled for a longer period of time,
covered more distance, and expended more calories than participants who cycled without
it. Similarly, Plante, Aldridge, Bogden, and Hanelin (2003) found that participants who
exercised in a virtual environment as opposed to exercising in a real environment
experienced less tiredness and reported more energy and enjoyment. Ijsselsteijn et al.
(2006) compared a low and high immersion environment and found that the high
immersion environment led to greater motivation. Van Schaik, Blake, Pernet, Spears, and
Fencott (2008) created a virtual exercise game for older adults and found that participants
underestimated the time spent exercising by 38%, indicating they were so engaged in the
VE they lost track of time and continued to exercise. A potential explanation is offered by
Huang, Tsai, Sung, Lin, and Chuang (2008), who performed an extensive analysis of
heart rate data while participants exercised in a virtual environment and found that VR
has physiological effects that may promote greater endurance for anaerobic activity.
These studies have demonstrated that virtual environments can create compelling and
effective environments for exercise-related stimuli.
42
CHAPTER EIGHT: MANIPULATING THEORETICAL CONSTRUCTS USING
VIRTUAL TECHNOLOGIES
Collectively, these studies demonstrate that virtual reality can keep participants
interested and motivated while exercising, potentially enhancing exertion, intensity, and
adherence, but they did not explore health-based theoretical explanations for these
effects. Social cognitive theory provides a useful framework for the development of
virtual health treatments. As Bandura (2002) noted, ―symbolic modeling lends itself
readily for society-wide applications through creative use of the electronic media‖ (p.
12). Indeed, virtual humans are useful models as they can be manipulated to portray a
range of desirable behaviors that may be difficult to enact in the real world. For example,
it is simpler to have a virtual human depict a skill at multiple levels of proficiency as
opposed to finding models ranging in their abilities from novice to expert.
One challenge in creating an effective stimulus is finding the model with the most
similarity to the target. According to social cognitive theory, observing a model similar to
the self should cause greater learning than observing a dissimilar model (Bandura, 1977,
2001). One possible explanation is that greater identification with a model leads to more
learning because it is easier to visualize the self in the place of the model. In a sense, high
identification may lead to a sort of embodiment wherein the observer really feels as if he
or she is having the same experience as the model. If the model is in fact a representation
of the self, this feeling may be even stronger, which may lead to a greater likelihood of
performing the behavior.
In current mass media campaigns, when modeling stimuli are created, the
maximum level of physical similarity is limited to matching the model to the observer on
43
categorical variables, such as sex and age (Singhal & Rogers, 2002). Given the strength
of model similarity in causing attitude and behavior change (Stotland, 1969), exceeding
these categorical variables may be beneficial. Virtual humans can portray the highest
level of similarity, including the same sex and age, but also the same skill level or
emotional state as the individual. Such similarities can help people develop feelings of
identification with and empathy toward virtual humans (Gilliath et al., 2008; Gratch &
Marsella, 2005), increasing their effectiveness as models and persuasive agents.
With previous efforts, a mirror or a videorecording of the self could capture the
model with the highest level of similarity (e.g., Dowrick, 1999). However, those
technologies have severe limitations because self-models are constrained by a person‘s
own skill to perform a given behavior. For example, the self could model a tennis serve,
but if one is a beginner, the ideal behavior of an ace serve may be too difficult to perform.
A professional baseball player could model a home run swing, but this may hamper the
individual‘s feelings of identification with the model because a Major League player is
not maximally similar to the self.
Another issue is that it may not be possible to model particular consequences.
Some consequences are simply outside the scope of model reenactment due to financial
feasibility, such as recreating the rewarding experience of a venue filled with cheering
fans. Others may be harmful, such as having a model smoke a cigarette to demonstrate
the punishment of a hacking cough. Although a ―before‖ and ―after‖ photograph of a
model can indicate consequences, long-term effects, such as the gradual damage of sun
exposure, are difficult to capture incrementally. Also, it is impossible to use the self as a
model because one cannot be transported into the future to observe the potential
44
outcomes. These issues may be resolved, however, with the incorporation of transformed
social interaction.
Transformed Social Interaction
The unique nature of virtual environments also led to the discovery of new
theoretical constructs. Virtual technologies enable us to modify interpersonal
communication in novel ways that we could not achieve in the real world, resulting in
transformed social interaction (TSI; Bailenson, Beall, Loomis, Blascovich, & Turk,
2004, 2005). According to Bailenson, Beall, Loomis, et al. (2004), TSI presents
advantages over traditional forms of communication in three realms. First, TSI presents
users with the opportunity to enhance their normal perceptual abilities (Bailenson &
Beall, 2006). For example, participants might be able to see other participants‘ names,
affiliations, or other relevant personal information hovering over their avatars.
Participants can also view an environment from different points in the room through
multilateral perspective taking. Second, VEs also enable manipulations of the context of
the interaction including time and space (Bailenson & Beall, 2006); participants may
choose to ―rewind‖ a conversation to hear part of it again, or ―pause‖ while they collect
their thoughts. Third, and perhaps the most fruitful realm for TSI research, is controlling
self-representation, namely ―decoupling the rendered appearance of behaviors of avatars
from the human driving the avatar‖ (Bailenson & Beall, 2006, p. 3). For example, identity
capture entails obtaining the participant‘s image and using software to morph it with
other individuals‘ images. Blending the two representations gives the other individual
some of the more familiar features of the self; the resulting similarity and familiarity
breeds more liking of this individual (Bailenson, Garland, Iyengar, & Yee, 2006).
45
One method of transforming the self-representation involves the use of a virtual
human that is photorealistically similar to the physical self but behaves independently of
the self. Recently, technologies have been developed to create virtual humans that bear
strong resemblance to individuals (Bailenson, Beall, Blascovich, & Rex, 2004; Bailenson
et al., 2008). Through the use of digital photographs and head-modeling software, an
individual‘s visage may be replicated in the virtual world. Although this transference is
not flawless, it creates relatively accurate models of the human form, and the striking
similarity gives these virtual representations of the self (VRSs) great potential as stimuli
in virtual realms. See Figure 2 for an example.
Figure 2. Constructing the virtual self. Photographs of an individual (top row) can be used to
create a realistic virtual model of the individual‘s head (bottom row).
46
The use of a virtual representation of the self (VRS) as a model has many
advantages over traditional modeling efforts. First, the VRS is maximally similar to the
self, so issues of identification with a model are overcome. The VRS can be matched on
any relevant quality to the real self. Second, compared to watching oneself on video, the
VRS is capable of performing activities at a higher level of proficiency than the physical
self, and the degree of skill can be modified as a continuous variable to present the
optimal model for each stage of learning. Third, the VRS is also capable of
transformation that the real self cannot easily achieve. For example, an obese man might
have difficulty envisioning himself thinner, or a thin man might not be able to fathom
gaining muscle mass. In the virtual world, these rewards can be portrayed as the VRS can
be altered to represent different levels of attainable or ideal body states. In sum, virtual
representations of the self have unique affordances that may make them more potent and
effective models than those used in traditional health behavior change efforts. The
following pretests examine how social cognitive theory can inform the use of VRSs as
models for health behavior change.
47
CHAPTER NINE: PRETESTS
Five pretests1 were conducted to explore the utility of virtual selves to model
health behaviors. In Pretest 1, vicarious reinforcement was manipulated to determine if a
virtual self experiencing the rewards and punishments of exercise was sufficient to
encourage imitation of the exercise behavior. In Pretest 2, rewards and punishments were
examined separately, and virtual selves were compared to virtual others on model
effectiveness. In Pretest 3, longer-term effects of exercising virtual selves and virtual
others were examined, and participants‘ real world exercise was measured 24 hours after
exposure. In Pretest 4, physiological measures were implemented to see if there was a
difference in bodily response to virtual selves and virtual others. In Pretest 5, virtual
selves were incorporated in a new domain, eating, and the role of presence was
examined.
Samples
All studies featured convenience samples drawn from the population of a midsized West Coast University. All participants were given class credit or paid $10-$15 for
their participation.
Apparatus
All studies featured the same general apparatus. Participants were placed in a
fully immersive virtual environment. They donned a head-mounted display (HMD)
through which they were able to view the stimulus. The HMD was an nVisor SX with
dual 1280 horizontal by 1024 vertical pixel resolution panels. The display presented a
visual field subtending approximately 50 degrees horizontally by 38 degrees vertically.
48
Stereoscopic images were rendered by a 1900 MHz Pentium computer with an NVIDIA
GeForce 6600 graphics card and were updated at an average frame rate of 60 Hz.
Sensing equipment tracked users‘ motions (e.g., turning their heads) so that a
realistic visual depiction of the environment could be updated constantly based on their
movements. In all studies except Pretests 4 and 5, the participant‘s movements along the
x, y, and z planes were tracked using WorldViz Precision Position Tracker optical
tracking system. Participants‘ head movements were tracked by a three-axis orientation
sensing system (Intersense IS250 with an update rate of 150 Hz) and used to continuously
update the simulated viewpoint. The system latency, or delay between the participant‘s
movement and the resulting update in the HMD, was no greater than 80 ms. Vizard 3.0
software was used to assimilate tracking and rendering.
Pretest 1: Vicarious Reinforcement of Virtual Selves
Overview and Hypotheses
This study investigated the use of virtual selves as exercise models in a virtual
environment. We hypothesized that seeing a VRS being rewarded for the participant‘s
physical exercise (through apparent weight loss) and punished for the participant‘s
inactivity (through apparent weight gain), as opposed to seeing an unchanging VRS or no
virtual human, would cause participants to engage in more voluntary exercise (H1). We
included an unchanging VRS as a control to determine whether it was vicarious
reinforcement of the self that induced the participant‘s exercise; the condition without a
virtual human was included to ensure that the effect was not merely seeing the self that
instigated exercise.
49
Procedure
The sample (N = 63) consisted of 31 women and 32 men aged 18 to 29
(M = 20.28, SD = 1.70).2
A between-subjects design was employed for this experiment. Participants were
randomly assigned to one of three Conditions: Reinforcement (n = 22), No Change
(n = 22), or No Virtual Human (n = 19).
Participants had their photographs taken with a digital camera for a presumably
unrelated study. Approximately six weeks after the photo session, participants were
solicited for the current study. Thus, all participants, regardless of condition, participated
in the photo session and had their virtual head models constructed. For the VRS
conditions, these heads were affixed to a sex-appropriate generic human body. Modeling
was limited to the head as modeling the physical body was beyond the human and
technological resources allotted to these studies.
Participants in all three conditions were provided with the same fitness prompt:
One of the greatest health issues facing Americans is physical inactivity. Exercise
is essential to maintaining a healthy body. A lack of exercise can lead to several
health problems, including obesity and cardiovascular disease. According to the
Surgeon General, people need at least 30 minutes of physical activity a day in
order to maintain their weight.
Next, participants were provided with a printout portraying a combination arm and
shoulder exercise. The research assistant demonstrated the exercise once, pausing to
describe each movement. Participants were handed two-pound weights, and they
performed the exercise slowly so that the research assistant could correct the motion as
needed before the treatment commenced. Then, the participant was immersed in the
virtual environment.
50
The experiment was structured in three phases that were consistent across
conditions. In the first phase, participants performed three sets of 12 exercises each. In
the second phase, participants stood still for two minutes. The third phase was voluntary:
participants were told they could stay in the virtual environment and, if they wished,
exercise, or they could end the experiment.
Although the phases were the same across conditions, there were differences in
what the participants observed in each condition. In the No Virtual Human condition,
participants saw nothing but an empty virtual room. In the other two conditions,
participants viewed their VRSs from the third person, consistent with previous research
involving VRSs (Bailenson et al., 2008). That is, the participant did not embody the
virtual self; the VRS was standing in the room facing the participant. In the No Change
condition, participants saw their own VRSs at an approximately average weight and body
shape in the virtual room, but their virtual bodies did not change for the duration of the
experiment. In the Reinforcement condition, the VRS started at the same average body
weight as in the No Change condition, but participants saw their VRSs appear to lose
weight as participants physically exercised or gain weight as they remained inactive. In
the first phase, for each repetition of the exercise, the VRS was scaled down 1% on the
X-axis, slimming the VRS by narrowing the body. In the second phase of mandatory
inactivity, for each 3 seconds the slimmed-down VRS was scaled up 1% on the X-axis,
widening the body until it was overweight. For the third phase, the VRS was reset to the
initial, average weight, and then the VRS appeared to lose or gain weight in accordance
with the participant‘s exercise or inactivity by the same percentages as the first two
phases. Figure 3 shows a VRS that has lost and gained weight.
51
Figure 3. Example stimuli for Pretest 1. As the participant exercised (left), the avatar got slimmer
(middle). If the participant did not exercise, the avatar gained weight (right).
The dependent variable was exercise repetitions. A research assistant counted
each exercise repetition participants performed during the voluntary third phase of the
experiment and recorded each with a keystroke. Exercise repetitions ranged from 0 to 51
(M = 6.38; SD = 13.76).
Results and Discussion
To test H1, we ran a one-way ANOVA with condition as the independent variable
and exercise repetitions as the dependent variable. There were significant differences
between the treatment groups, F(2, 60) = 11.08, p < .0005, partial η2 = .27. A follow-up
Fisher‘s LSD test revealed that the only significant differences were that participants in
the Reinforcement condition exercised more frequently (M = 16.04, SD = 19.35) than
participants in the No Change (M = 1.68, SD = 5.23) and No Virtual Human (M = 0.63,
SD = 2.75) conditions, as is seen in Figure 4.
52
Figure 4. Results of Pretest 1. The vicariously reinforced avatar encouraged significantly more
exercise than an unchanging avatar or no avatar.
In support of H1, these findings indicated that vicarious reinforcement was
successful: seeing the VRS rewarded for performing an exercise behavior and then
punished for not performing it encouraged exercise behavior. Simply being immersed in
a virtual environment or seeing the static VRS in a virtual environment while exercising
was not sufficient. Observing the VRS losing weight in accordance with one‘s physical
exercise and seeing the VRS gain weight due to physical inactivity effectively
encouraged participants to engage in exercise.
Pretest 2: Reward and Punishment of Virtual Selves and Virtual Others
Overview and Hypotheses
This study was designed to determine if either portrayals of reward or punishment
were more effective in increasing exercise behavior. Previous studies on exercise
motivation have indicated that the promise of reward often leads to increased exertion
53
(Buckworth, Lee, Regan, Schneider, & DiClemente, 2007). Thus, it was hypothesized
that those in the reward conditions would exercise more than those in the punishment
conditions (H2). Additionally, this study addressed the concept of identification through
similarity with the model. We used virtual humans that looked similar or dissimilar to the
self as exercise models to determine whether the virtual self was essential, or if any
virtual model could achieve the same effects. Because model similarity might promote
identification and thus performance, we hypothesized that the VRS would be a more
effective model than a VRO (H3).
Procedure
An initial sample of 60 was obtained; due to technological failure, seven
participants were dropped from analyses. The final sample (N = 53) included 21 women
and 32 men aged 18 to 55 (M = 20.54, SD = 5.81).
A 2 x 2 between-subjects design was employed for this experiment. Participants
were randomly assigned to one of four Conditions: VRS-Reward (n = 14), VRSPunishment (n = 12), VRO-Reward (n = 14), or VRO-Punishment (n = 13).
The same photograph and head-modeling procedure used in the first study was
employed in Study 2. For the VRS conditions, participants‘ heads were affixed to a sexappropriate generic human body. For the VRO conditions, the virtual human featured an
unknown person‘s head of the same sex and approximately the same age that was
selected randomly for each participant from a pool of past experimental participants not
involved in the current study.
Participants in all three conditions were provided with the same fitness prompt as
Study 1. Participants in the Reward conditions were informed their avatars would be
54
losing weight in accordance with their exercise; those in the Punishment conditions were
informed their avatars would be gaining weight in accordance with their inactivity. Next,
the research assistant demonstrated the exercise for this study, marching in place. A
different exercise was chosen to increase generalizability. Participants were instructed to
lift their right knee to waist level and then return their right foot to the ground; then they
repeated the action with their left leg. A right leg, left leg sequence was counted as one
repetition. The research assistant demonstrated the exercise once slowly and then once at
a normal pace to show how the repetitions would be counted. Then, the participant was
immersed in the virtual environment.
All three conditions were structured in three phases similar to the previous study.
In the first phase, participants performed three sets of 20 exercises each. In the second
phase, participants were asked to stand as still as possible for two minutes. Participants
were told the third phase was voluntary: they could stay in the virtual environment and
exercise if they wished, or they could end the experiment. In all conditions, similar to the
first study, participants saw their virtual representations from the third-person
perspective. As participants exercised in the physical world, their virtual selves also
exercised. In the first phase of the VRS-Reward condition, participants saw the VRS lose
weight as they performed the mandatory exercises. In the second phase, the VRS
remained inactive, but no consequences were shown for inactivity. In the third phase, if
the participant chose to exercise, the VRS exercised and lost weight. If the participant did
not exercise, no consequences were shown. The VRO-Reward condition was similar
except that participants saw a virtual other instead of the virtual self. Thus, in the Reward
conditions, no punishment was shown for inactivity; only reward was shown for exercise.
55
In the VRS-Punishment condition, the VRS did not lose weight in the first phase as
participants exercised. In the second phase, as the participant was inactive, the VRS
gained weight. In the third phase, if the participant chose to exercise, the VRS exercised
but did not change; if the participant was inactive, the VRS gained weight. The VROPunishment condition was the same except that a virtual other was the model. Thus, in
the Punishment conditions, no reward was shown for the desired behavior, only
punishment for the undesired behavior.
Measures
Exercise repetitions. As in Study 1, the dependent variable of interest was
exercise repetitions. A research assistant counted each exercise repetition participants
performed during the voluntary third phase of the experiment and recorded each with a
keystroke. Forty-three participants (81%) chose to exercise during the third phase; the
number of repetitions performed ranged from 0 to 215 (M = 44.87; SD = 51.05).
Virtual human resemblance. In order to ensure that the self-other manipulation
was successful, participants were asked to indicate on a fully-labeled five-point scale the
degree to which they felt the virtual human resembled them (1 = Definitely did not look
like me at all; 5 = Definitely looked a lot like me; M = 2.35, SD = 1.06). Participants in
the VRS conditions (M = 2.71, SD = .98) reported that the virtual human looked more
like them than participants in the VRO conditions (M = 1.96, SD = 1.02), t(53) = 2.79,
p < .01, Cohen‘s d = .75, confirming that the self-other manipulation was successful.
Results and Discussion
A 2 x 2 ANOVA was performed to determine the effect of conditions on exercise
repetitions. Regarding main effects, H2 was not supported. There was no difference
56
between Reward (M = 49.07, SD = 49.33) and Punishment (M = 40.14, SD = 53.52)
conditions, F(1, 49) = .38, p > .05, partial η2 = .01. The self-other hypothesis (H3) was
confirmed: those in the Self conditions performed significantly more exercises
(M = 62.23, SD = 64.17) than those in the Other conditions (M = 28.15, SD = 25.70),
F(1, 49) = 6.15, p < .02, partial η2 = .11. The interaction effect was not significant,
F(1, 49) = .14, p > .05, partial η2 = .00. Figure 5 illustrates these findings.
Figure 5. Results of Pretest 2. A main effect was found for seeing the virtual self experience
changes as opposed to a virtual other.
This study determined that the virtual self can be used as a model to encourage
exercise, whereas a virtual other is not sufficient. Using the virtual self, this study did not
detect any differences regarding whether the participant was rewarded for their exercise
or punished for not exercising: either manipulation stimulated exercise. The small sample
57
for this study may have prevented finding a difference between rewards or punishments,
however. Alternatively, the benefit of weight loss and the threat of weight gain as
depicted in this manipulation may have been qualitatively equivalent.
Pretest 3: Effects of Exercising Virtual Selves and Virtual Others After 24 Hours
Overview and Hypothesis
Similar to Pretest 2, this study explored the use of virtual humans that looked
similar and dissimilar to the self as exercise models in a virtual environment. Rather than
examining immediate effects inside the laboratory, this study focused on whether these
stimuli caused participants to exercise later in the physical world. A VRS exercising
should maximize identification, whereas a VRO matched solely on the categorical
variables of sex and age should provide a limited degree of identification. Thus, we
hypothesized that seeing a VRS running, as opposed to a VRO running or VRS loitering,
would increase participants‘ physical activity following exposure (H4).
Procedure
An initial sample of 75 was obtained; two participants failed to complete the
follow up survey and their data were excluded from the analyses, leaving a sample of
N = 73, including 50 women and 23 men aged 18 to 33 (M = 20.61, SD = 2.50).
A between-subjects design was employed. Participants were randomly assigned to
one of three Conditions: VRS-Running (n = 25), VRS-Loitering (n = 24), or VRO-Running
(n = 24). None of the participants from this study had participated in the first or second
study.
The same photograph and head-modeling procedure used in the first two studies
was employed in Study 3. In all three conditions, participants observed a virtual human
58
for 5 min 20 s. During this time, they engaged in a distracter task in which they focused
on a sequence of 20 numbers that flashed on the virtual human‘s chest for later recall.
The purpose of the distracter task, derived from previous work (Bailenson et al., 2003),
was to keep the participant visually attended to the virtual human as well as to mask the
experimental manipulation. In the VRS-Running condition, the virtual human was running
on a treadmill and featured the participant‘s face on a sex-appropriate generic virtual
human body. In the VRO-Running condition, the only difference was that the virtual
human featured another unknown person‘s face of the same sex and approximately the
same age that was selected randomly for each participant from a pool of past
experimental participants not involved in the current study. In the VRS-Loitering
condition, the virtual human was standing, shifting its weight, and occasionally crossing
its arms; the participant‘s face was featured on a sex-appropriate generic virtual human
body. Figure 6 provides examples of running and loitering virtual humans.
Figure 6. A running virtual human (A) and a loitering virtual human (B).
59
Twenty-four hours after the experiment ended, the researcher called all
participants to remind them to fill out the survey and emailed the survey link. Participants
who had indicated they would not be near a computer 24 hours following the experiment
(n = 9) had been provided with a sealed survey to complete on paper; later, they
submitted those responses electronically.
Measures
Virtual human resemblance. Participants were asked to rate on a fully-labeled
five-point scale the degree to which they felt the virtual human resembled them
(1 = Definitely did not look like me at all; 5 = Definitely looked a lot like me; M = 2.45,
SD = 0.94). This served as a manipulation check to ensure that participants felt they were
viewing a VRS as opposed to a VRO. A one-way ANOVA determined there were
significant differences between the three conditions, F(2, 72) = 9.56, p < .001, partial
η2 = 0.22. A follow-up Fisher‘s LSD test at α = .05 revealed that participants rated the
virtual humans in the VRO condition (M = 1.83, SD = 0.92) as resembling themselves
significantly less than participants in the VRS-running (M = 2.72, SD = 0.89) or VRSloitering (M = 2.79, SD = 0.72) conditions.
Paffenbarger Physical Activity Questionnaire. A modified version of the
Paffenbarger Physical Activity Questionnaire (PPAQ; Paffenbarger, Wing, & Hyde,
1978) was administered in the follow-up survey. Participants were asked to reflect on the
24 hours that had transpired since the experiment ended. They reported how many city
blocks (or miles) they had walked and how many flights of stairs they had climbed.
Participants were provided with a list of activities representing nine levels of MET
(metabolic equivalent, essentially a measure of calorie burn) activity ranging from
60
sleeping to vigorous activity. They were asked to indicate how much time (in 15-minute
increments) they had engaged in these or similar activities over the past 24 hours
(following Aadahl & Jørgensen, 2003). The items of interest were those that involved
physical activity, which included city blocks walked, flights of stairs climbed, and the
four highest levels of MET activity: 4.0 MET (bicycling and brisk walking), 5.0 MET
(carrying and loading items), 6.0 MET (aerobics or other health club exercise), and > 6.0
MET (intense exercise such as running or playing soccer).
Intent of study. Participants were asked to speculate on the intent of the study.
Both the researcher and an independent coder blind to experimental condition evaluated
the written open-ended responses for any mention of modeling or mimicking the virtual
human‘s activity. No participant correctly identified its purpose.
Results and Discussion
To address H4, the items that entailed physical activity were examined. Because
these items used different units of analysis, the variables were standardized and Z-scores
were derived. Next, a principal components analysis with Promax rotation was performed
on the six physical activity items. Three items loaded on the first factor, which explained
28% of the variance; two items loaded on the second factor, which explained 24% of the
variance. One item loaded equally on both factors and was consequently dropped from
further analyses. The items and factor loadings can be viewed in Table 1; Table 2 shows
the correlations between items.
61
Table 1
Factor Loadings for Physical Activity Items
Commute
Exercise
City blocks walked
.785
.061
Flights of stairs climbed
.582
.175
Biking, brisk walking (Activity at 4.0 MET )
.653
-.186
Aerobics, gym workout (Activity at 6.0 MET)
-.004
.757
Running, playing soccer (Activity at >6.0 MET)
.050
.748
Commute (Cronbach‘s α = .41)
Exercise (Cronbach‘s α = .31)
Factor 1, Commute, explains 27.81% of variance; Factor 2, Exercise, explains 23.85% of
variance.
Table 2
Correlation Matrix for Factor Items
C1 Blocks
C2 Stairs
C3 Biking
E1 Exercise
E2 Intense Ex
C1 Blocks
C2 Stairs
C3 Biking
E1 Exercise
E2 Intense Ex
--
.23*
.26*
.05
.02
--
.07
.01
.06
--
-.05
-.01
--
.18
--
* p < .05
The three items that loaded on the first factor (number of blocks walked, number
of flights of stairs climbed, and other activities at 4.0 MET, such as bicycle riding) were
termed Commute, as these activities generally reflect the activity required to get to or
62
from places and are likely related to commuting. The two items on the second factor,
exercise and intense exercise, were labeled Exercise. These items reflected activities
aside from everyday transportation that required time beyond the work day. Although the
Kaiser-Meyer-Olkin statistic for the factors was low (.51), it exceeded the threshold for
factor analysis (Kaiser, 1974). Additionally, the two factors parallel those identified by
Baecke, Burema, and Fritjers (1982).
The three conditions were compared on each of these two factors. A one-way
ANOVA found no significant differences between the VRS-Running (M = -0.14,
SD = 0.79), VRO-Running (M = 0.00, SD = 1.09), and VRS-Loitering (M = 0.15,
SD = 1.12) conditions for the Commute items, F(2, 70) = .48, p > .05, partial η2 = .01. In
contrast, an ANOVA revealed significant differences across conditions for the Exercise
items, F(2, 70) = 3.51, p < .05, partial η2 = 0.09. Follow-up LSD tests at α = .05 showed
that participants in the VRS-Running condition engaged in significantly more exercise
(M = 0.42, SD = 1.22) than participants in the VRO-Running (M = -0.21, SD = 0.62) and
VRS-Loitering (M = -0.22, SD = 0.95) conditions. When we look at the data in actual
minutes as opposed to factor scores, participants in the VRS-Running condition
(M = 142.2, SD = 146.02) engaged in over an hour more voluntary exercise independent
of commuting than participants in the VRO-Running (M = 66.88, SD = 65.00) and VRSLoitering (M = 61.88, SD = 94.80) conditions, as can be seen in Figure 7.
63
Figure 7. Results of Pretest 3. Although there were no differences in commuting, participants
who saw a VRS running exercised significantly more than other conditions.
These findings indicate that seeing one‘s VRS model an exercise behavior
stimulates the performance of exercise behavior in the individual. It was not merely
seeing a VRS that encouraged this behavior, as seeing the VRS loitering did not lead to
an increase in activity. Also, seeing a VRO exercise was not sufficient to encourage the
individual to exercise. Of course, this finding should be considered within its context,
among an active college population in a health-oriented area of the country.
It is important to note that these results are preliminary, as the factor analysis
showed a less than desirable K-M-O statistic. The correlations between items on the
scales were low, but this may be due to the zero-sum nature of activity over a twenty-four
hour period: for example, exercise at a very high intensity might displace the hours spent
exercising at a regular intensity. Thus, further examination of these factors is necessary.
64
Pretest 4: Physiological Responses to Virtual Selves and Others
Overview and Hypothesis
For this study, we wanted to address whether observing self or other models, or
exercising or loitering models, resulted in differing levels of physiological arousal.
Because the self-running model in the previous study led to more exercise, we
hypothesized that this model may yield greater physiological arousal than a loitering self,
a running other, or a loitering other (H5).
Apparatus
For this study, in addition to the virtual reality apparatus, equipment was used to
gather physiological data from participants. Physiological signals were measured,
amplified, and recorded using a Thought Technology ProComp Infiniti module linked to
a computer. Thought Technology‘s Infiniti software program coordinated the sampling
and storage of physiological data. Heart rate was recorded using a blood volume pulse
detection sensor placed on the index finger of the non-dominant hand. Data were initially
collected as interbeat intervals or milliseconds between consecutive R-spikes sampled at
a rate of 256 times per second. They were then edited and converted off-line to heart rate
in beats per minute. Skin conductance was recorded using standard Ag/AgCl electrodes
placed on the ring finger of the non-dominant hand. The signal was sampled at a rate of
32 times per second and converted to conductance values in microSeimens (uS). See
Figure 8 for a depiction of how participants were hooked up to the equipment.
65
Figure 8. The setup for physiological measures during immersion in the HMD.
Procedure
The sample (N = 22) consisted of 10 men and 12 women ranging in age from 18
to 29 (M = 21.64, SD = 3.84). Due to equipment failure and physiological nonresponse
(i.e., participants were not perspiring), skin conductance data was only viable for 10
participants (five men and five women).
Participants were advised to remain as still as possible during the treatment. Each
participant was presented with the four phases in a randomized order. The phases
included a VRS running, a VRS loitering, a VRO running, and a VRO loitering. During
each phase, participants witnessed an avatar (of either the self or another unknown person
of the same sex and approximately the same age) that was either running on a treadmill or
standing in place, occasionally shifting its weight or crossing its arms. Refer to Figure 6
for illustrations of a virtual representation running and loitering.
66
Physiological Data Processing
Heart rate. Before analyzing the heart rate data, we performed two preprocessing
steps to remove noise. First, we scanned the data for outliers, which we defined as points
that differed by more than three BPM from the median of their five closest neighbors. We
replaced such points with the corresponding medians. Next, we smoothed the data using a
―moving average,‖ replacing each point with the mean of its five closest neighbors.
We measured heart rate elevation in terms of an increase over a baseline. As the
participants‘ heart rates did not always return to a resting state before each phase of the
experiment, we used a separate baseline for each condition. The baselines were collected
by taking the average heart rate over a 5-second interval immediately preceding each
phase.
Each condition lasted 160 seconds; however, we found that the participants
tended to become habituated after about two minutes, causing their heart rates to return to
their baseline levels. Consequently, we considered only the first 105 seconds of each
phase. To calculate the percent increase in heart rate for each condition, we subtracted the
baseline from the participant‘s average heart rate during this 105-second interval, and
then divided by the baseline.
Although subtracting the baseline heart rates allowed us to take into account the
differences in resting heart rates between subjects, it did not capture differences in
variability. For example, a 5% increase in heart rate for a subject with low variability
may be more significant than a 10% increase for a subject with high variability. To
address this issue, we applied vector normalization to the four data points representing a
67
subject‘s percent increase in each phase. The vector-normalized data was used for
analysis.
Skin conductance. The skin conductance data was characterized by rapid spikes of
varying amplitudes. Such a spike is referred to as a skin conductance response (SCR).
From this data we extracted several features for each experimental phase, including the
number of small, medium, and large SCRs, and the mean and maximum amplitude of all
SCRs. The mean amplitudes were used for analysis.
Results and Discussion
No differences were found in heart rate across conditions, F(3, 63) = 1.84,
p > .05, partial η2 = .08. The means suggested that the VRS-Running (M = .14, SD = .48)
and VRO-Loitering (M = .15, SD = .50) phases stimulated an increase in heart rate
whereas the VRS-Loitering (M = -.11, SD = .57) and VRO-Running (M = .04, SD = .43)
stimulated a decrease, although the differences between these phases were not significant.
Results indicated a significant difference in skin conductance across conditions,
F(3, 27) = 3.50, p < .05, partial η2 = .28. Planned contrasts in the form of paired t-tests
revealed that participants demonstrated significantly greater skin conductance during the
VRS-Running phase (M = .29, SD = .49) than during in the VRS-Loitering phase
(M = -0.30, SD = .41; t(9) = 3.27, p < .01, Cohen‘s d = 1.31) and the difference from the
VRO-Running phase bordered on significance (M = -0.29, SD = .58; t(9) = 1.91, p = .088,
Cohen‘s d = 1.08). Interestingly, there was no difference between VRS-Running and
VRO-Loitering (M = .31, SD = .52), t(9) = .06, p > .05, Cohen‘s d = .04. Additional
analyses revealed that participants experienced greater skin conductance in the VROLoitering phase than during the VRS-Loitering [t(9) = 2.40, p < .05, Cohen‘s d = 1.30] or
68
VRO-Running phases [t(9) = 2.26, p = .05, Cohen‘s d = 1.09]. Results can be seen in
Figure 9.
Figure 9. Results of Pretest 4. Participants experienced greater arousal to a VRS running and a
VRO loitering than a VRS loitering and a VRO running.
Participants did not show a difference in heart rate across phases. This may be an
artifact of the extensive data-cleaning that took place. The equipment can occasionally
present erratic numbers (e.g., heart rates doubling and reverting within 1/32 of a second)
and analyzing this data requires many judgments on determining the accurate readings. It
is possible that our implementation of a moving average, for example, washed out some
variance that would have yielded differences.
Participants experienced greater skin conductance in response to running selves
and loitering others than to running others or loitering selves. It may be that running
69
selves are stimulating in that participants feel as if they are actually engaged in the
exercise. Alternatively, it may serve as a cue that summons the feelings of previous
experiences with exercise, causing the body to respond in a similar physiological manner.
The finding that the other loitering also caused greater physiological arousal is somewhat
puzzling. However, it is possible that, in contrast to the running other (which had a clear
purpose), the loitering other made participants uncomfortable and made them feel they
were being watched by a stranger. The loitering self may not have resulted in discomfort
and the resultant peak in arousal because it is not dissimilar to looking in the mirror, a
common experience for most.
These results may offer some insight on the results of Pretest 3, wherein a running
self promoted more exercise in the 24 hours following the experiment than a loitering self
or a running other. Participants experienced greater physiological arousal at seeing the
virtual self run than the self loitering or another person running, which may have served
as a physical impetus to exercise after the experiment. Choosing to exercise after
experiencing greater arousal may have been a form of excitation transfer. Future research
may expose participants to these stimuli and then present them with opportunities to
address this arousal in the laboratory immediately afterwards (e.g., through exercise or
other physical tasks) to observe and compare differences in subsequent behavior.
Pretest 5: Presence and the Imitation of Eating Behaviors
Overview and Hypothesis
This experiment was designed to examine the effect of experiencing the virtual
self eating on subsequent physical eating behavior. Participants were exposed to a
stimulus in which they observed the virtual self eating. To externally manipulate factors
70
contributing to presence, their virtual bodies either changed or did not change in
accordance with the modeled behavior.3 Participants‘ internal, subjective experience of
presence was assessed with a memory task that took place in the virtual environment as
well as questionnaire items immediately following the treatment. While completing the
questionnaire, participants were seated at a computer with a bowl of chocolates and given
the opportunity to eat candy.
Presence has been shown to bolster the likelihood of the transfer of virtual
experiences to real world behaviors. Thus, it was hypothesized that those who experience
high levels of psychological presence will demonstrate more imitative eating behavior
than those who experience low levels of presence (H6).
Procedure
Four participants were dropped from the initial sample (N = 73) due to technical
failure during the experiment or because they reported feeling ill that day. The final
sample (N = 69) consisted of 32 men and 37 women who ranged in age from 18 to 29
(M = 20.20, SD = 1.55).
A between-subjects design was employed. Participants were randomly assigned to
one of two Conditions: Change (n = 32) or No Change (n = 37). At the beginning of the
quarter, participants had their photographs taken with a digital camera for a presumably
unrelated study. Approximately one month after the photo session, participants were
solicited for the current study. Thus, all participants, regardless of condition, participated
in the photo session and had their virtual head models constructed.
When participants entered the lab, they were instructed as such:
While inside the virtual world, you are going to observe your virtual self. Your
virtual self will be presented with food and commence eating.
71
In the changing conditions, participants were also told:
You will then see yourself experience the consequences of dietary choices: if your
Avatar makes healthy decisions, it will lose weight. If your avatar makes
unhealthy decisions, your avatar will gain weight.
All participants were also told:
As your avatar eats, you will also be engaging in a memory task. A sequence of
numbers will appear. Your goal is to remember as many of those numbers as
possible. After the sequence is finished, you‘ll be asked to recall those numbers.
Participants were seated at a table and outfitted in the HMD. In the virtual world,
participants saw themselves positioned between two bowls, one full of carrots and one
full of candy. Both bowls were labeled, and a package of each item was positioned near
the bowl (a bag of carrots and a package of Reese‘s Cups.) The experimenter pressed a
key to commence the eating behavior. A computer algorithm randomly determined
whether the avatar would eat carrots or candy first. After three minutes, the eating
animation would stop and the avatar would then begin to eat the other food for three
minutes. In the changing body conditions, the avatar lost weight when it ate carrots and
gained weight when it ate candy. In the unchanging body conditions, the avatar did not
appear to lose or gain weight while eating. See Figure 10 for an illustration of the
changing body condition.
72
Figure 10. Example stimuli from Pretest 5. In the changing condition, participants saw a virtual
self slim as it ate carrots (top row) and gain weight as it ate Reese‘s Cups (bottom row).
After exposure, participants were led to a computer and asked to complete the
survey items. A bowl of chocolate candy (Hershey‘s Kisses and Rolos instead of the
Reese‘s Cups featured in the stimulus) was placed next to the computer and participants
were told they could help themselves if they wished. To ensure that participants did not
feel as if they were being observed, the experimenter made an excuse to step outside the
room and told participants to retrieve him or her when the survey was completed.
Measurement
Numbers identified: presence proxy. Participants were presented with ten sets of
numbers and asked which of the numbers they recalled seeing while in the virtual world.
The total of numbers identified ranged from 5 to 10 (M = 8.59, SD = 1.33). The purpose
73
of the distracter task, employed in Pretest 3 as well as previous VE studies (Bailenson et
al., 2003), was to keep the participant visually attended to the virtual human as well as to
mask the experimental manipulation. This task also serves as a more objective measure of
presence than subjective questionnaire items, which have been criticized for their validity
(Slater, 2004). Because of individuals‘ limited cognitive resources, the more accurate
participants‘ memory is for the numbers, the less attention they were paying to the
stimulus and the virtual environment. In this sense, memory should be a proxy for
presence.
Self-reported presence. A subjective measure of presence was also used. Eleven
items were used to assess participants‘ experience of presence while immersed in the
virtual world. These items were culled from several sources (Bailenson & Yee, 2007;
Nowak & Biocca, 2003; Witmer & Singer, 1998). Participants indicated on a 5-point
scale (1 = Not at all; 5 = Extremely) the degree to which they felt present in the virtual
environment. Because the items were derived from multiple sources, a factor analysis
was conducted; the results indicated that there was only one factor, and thus all items
were combined to create the scale. Responses were averaged; scores ranged from 1.20 to
4.20 (M = 2.52, SD = .64). A Cronbach‘s alpha of α = .88 was achieved.
Number of candies eaten. The experimenter counted the number of candies
remaining in the bowl after the participant left to determine how many were eaten.
Participants ate between 0 and 8 candies (M = 1.30, SD = 2.28).
Open-ended response. Participants were asked to respond to the following
prompt: ―Please describe any reactions you have to seeing this representation of yourself.
How did it make you feel?‖
74
Results and Discussion
Because of previously identified sex differences in terms of presence (Nowak,
Krcmar, & Farrar, 2008) as well as eating attitudes, norms, and behaviors (Baker, Little,
& Brownell, 2003; Rosen, Silberg, & Gross, 1988), sex was entered as a factor in the
analyses. Thus, 2 by 2 ANOVAs were run for the manipulation checks as well as the
hypothesis test. All assumptions for the ANOVA were met for reported tests unless
otherwise noted.
Manipulation checks. For memory, there was a main effect of Change,
F(1, 65) = 6.07, p < .05, partial η2 = .09. Participants in changing body conditions
(M = 8.16, SD = 1.61) identified significantly fewer numbers than those in unchanging
body conditions (M = 8.97, SD = .90). Neither the main effect for Sex, F(1, 65) = .14,
p > .05, partial η2 = .00, or the interaction effect, F(1, 65) = .28, p > .05, partial η2 = .00,
were significant.
For self-reported presence, the main effect for Change was significant,
F(1, 65) = 4.83, p < .05, partial η2 = .07. Participants in the changing body condition
(M = 2.68, SD = .56) reported more presence than those in the unchanging body
condition (M = 2.38, SD = .68). The main effect for Sex also bordered on significance,
F(1, 65) = 3.53, p = .07, partial η2 = .05. In line with the findings of Nowak et al. (2008),
there was a trend for men (M = 2.65, SD = .71) to self-report more presence than women
(M = 2.41, SD = .55). The interaction effect was not significant, F(1, 65) = 1.44, p > .05,
partial η2 = .02.
Hypothesis. In order to examine the role of self-reported presence on modeling
the eating behavior (H6), we performed a median split to separate participants into low
75
and high presence groups. Those scoring at or below the median (Mdn = 2.40) were
categorized as low presence (n = 34), whereas those scoring above the median (n = 35)
were categorized as high presence. Sex was retained as a variable in the analyses.
To confirm the median split of subjective responses continued to correspond to
memory differences, a 2 by 2 ANOVA revealed a main effect for self-reported presence
on number identification, F(1, 65) = 5.27, p < .05, partial η2 = .08. Those in the low
presence group (M = 8.94, SD = .95) identified more numbers than those in the high
presence group (M = 8.26, SD = 1.56).
An examination of the number of candies consumed revealed some outliers. Thus,
data for participants who consumed more than 3 candies (n = 9) were wisterized to 3. A 2
by 2 ANOVA on number of candies eaten revealed no main effect for presence,
F(1, 53) = .00, p > .05, partial η2 = .00, and no main effect for sex, F(1, 53) = .17,
p > .05, partial η2 = .00.4 The interaction effect, however, was significant,
F(1, 53) = 4.95, p < .05, partial η2 = .09. It should be noted that the Levene‘s statistic for
this test was significant, indicating unequal variances. The ANOVA, however, is a robust
test, and the cells were all relatively equal in sample size. Within-sex follow-up tests
revealed that low presence females (M = 2.00, SD = 2.95) ate more candies than high
presence females (M = .36, SD = .94), t(16.41) = 2.07, p = .055. High presence males
(M = 2.06, SD = 2.79) ate significantly more candies than low presence males (M = .50,
SD = .67), t(17.28) = 2.16, p < .05. Figure 11 depicts these findings.
76
Figure 11. Results of Pretest 5. High presence men consumed more candy than low presence
men, whereas high presence women consumed less candy than low presence women.
A 2 by 2 ANOVA (Change by Sex) was run to ensure that subjective presence
was driving this effect rather than the environmental manipulations. Neither the main
effect for Change, F(1, 53) = .12, p > .05, partial η2 = .00, nor the interaction effect,
F(1, 53) = .07, p > .05, partial η2 = .00, were significant.
This study revealed that viewing one‘s virtual body change while eating limited
the ability to recall numbers. Body change was also related to the self-reported
experience of presence: seeing changes kept participants more engaged and they reported
higher levels of presence than those in the unchanging condition. When participants were
divided into low and high presence groups based on their subjective self-report, it was
discovered that those in the high presence group identified fewer numbers than those in
the low presence group. A self-reported Presence by Sex effect was also found on the
77
number of candies consumed; low presence females and high presence males ate more
candy than high presence females and low presence males.
Justification for Dissertation Studies
Although these pretests have provided compelling evidence for the effectiveness
of virtual treatments rooted in social cognitive theory, future research is needed to further
explicate the mechanisms behind these effects and determine the utility of these
technologies. The studies were designed to achieve two goals. Study 1 was designed to
see if virtual models that varied on degree of similarity (self, same race as self, or
different race than self) evoked different levels of identification, self-efficacy, and
exercise repetitions. Study 2 was designed to consider pre-existing self-efficacy regarding
exercise behavior and compare the effectiveness of virtual stimuli with another common
form of exercise motivation, mental imagery.
A major shortcoming of the pretests is that the role of self-efficacy was not
determined before or after the treatments were administered. Believing that one is able to
exercise, lose weight, and maintain a healthy routine has been shown to be a significant
contributor to adherence to diet and fitness plans (Bandura, 1997). It is possible that the
treatments featured in these studies either enhanced or diminished feelings of selfefficacy, which then affected participants‘ exercise behaviors. Another possibility is that
participants had pre-existing levels of self-efficacy regarding the modeled behavior, and
those beliefs determined how (or if) the treatment was successful in promoting healthful
attitudes and behavior.
The pretests compared the effectiveness of different virtual stimuli. However, it
could be that the virtual environment is not necessary to have participants experience the
78
message. Perhaps a prompt could inspire the individual to imagine the scene effectively
enough to yield the same behavioral effects.
The dissertation studies were intended to answer these questions. Study 1 was
designed to examine variation in the degree of identification with a model by comparing
the virtual self with a similar other and a dissimilar other. This study was designed to
determine whether these models would have differential effects on self-efficacy as well
as exercise behavior. Study 2 considered whether a fully immersive virtual environment
was necessary to yield imitation of exercising self-representations. Participants either
observed their virtual selves successfully completing exercises in the VE or were asked to
mentally imagine the same scene. This study also contrasted participants who had high
and low pre-existing levels of self-efficacy regarding the modeled exercise behavior to
see if they would be differentially affected by the two treatments.
79
CHAPTER TEN: DISSERTATION STUDY ONE: DEGREES OF IDENTIFICATION
WITH VIRTUAL SELVES AND OTHERS
This study was designed to add further understanding to the distinction between
virtual selves and virtual others. None of the pretests systematically varied the level of
similarity between the self and the virtual other. Thus, this experiment was designed to
compare three distinct levels of resemblance—a virtual self, a similar virtual other of the
same race, and a dissimilar virtual other of a different race—to see whether different
levels of identification impacted self-efficacy and exercise behaviors.
This study sought to examine, first and foremost, whether virtual self-models
could be used to enhance self-efficacy regarding a specific exercise behavior, and
whether these models would be more effective than similar or dissimilar models. Second,
this study sought to replicate previous findings regarding the use of virtual selves to
promote exercise behaviors. Because the avatars used in this study only vary in the face
(and hands, which match the skin tone of the face), the extent to which resemblance
could be manipulated was limited to very basic demographic variables. Given that sex
has had some mixed findings (i.e., females are more likely to imitate males than males
are to imitate females, see Bussey & Perry, 1982, for a discussion) and college-aged
students tend to have negative conceptions towards older people (Greenberg, Schimel, &
Martens, 2002; Rupp, Vodanovich, & Credé, 2005), race/ethnicity was determined to be
the only viable alternative. Several studies have examined race matching and
mismatching in the context of modeling and found that participants identify with racematched models more than race mismatched models (e.g., Anderson & McMillion, 1995;
Appiah, 2001; Kalichman et al., 1993; Pitts, Whalen, O‘Keefe, & Murray, 1989).
80
The third goal of this study was to examine other variables which may explain
previous findings. First, previous studies only examined resemblance as a measure of
identification with the model. However, identification is a psychological process that
entails more than physical likeness, and so a measure was added to determine
participants‘ experiences of identification (Cohen, 2001). In Pretest 3, participants saw a
specific form of exercise behavior, running, yet they reported an increase in various types
of health behaviors beyond running. It could be that these stimuli cue a general exercise
efficacy schema, not just to the specific behavior, which is why exercise in general
increased. Thus, general exercise self-efficacy was also considered. Finally, in Pretest 5,
presence was shown to play a role in the subsequent imitation of behavior, so this
measure was also included. The following hypotheses and research questions were
proposed:
H1: Participants in the Self condition will report higher levels of identification
with their avatar than those in the Similar Race condition, who will report
higher levels of identification than those in the Dissimilar Race condition.
H2: Participants in the Self condition will report higher levels of specific exercise
self-efficacy than those in the Similar Race condition, who will report
higher levels of specific self-efficacy than those in the Dissimilar Race
condition.
H3: Participants in the Self condition will complete more exercise repetitions than
those in the Similar Race condition, who will complete more exercise
repetitions than those in the Dissimilar Race condition.
81
RQ1: Will there be a difference in general exercise self-efficacy across
conditions?
RQ2: Will there be a difference in presence across conditions?
Method
Sample
A convenience sample was obtained from the student population of a mediumsized West Coast university. The initial sample consisted of 80 participants, but four were
dropped due to technological difficulties and one was dropped due to failure to follow
instructions. The final sample (N = 75) consisted of 38 women and 37 men aged 18 to 44
(M = 20.95, SD = 3.50). Participants self-reported their race/ethnicity as
Caucasian/European-American/White (44%; n = 33); African/African-American/Black
(12%; n = 9); Latino/a (8%; n = 6); Asian/Asian-American (22.7%; n = 17); American
Indian (1.3%; n = 1); and Other or multiple races/ethnicities (10.7%; n = 8).
Design
A between-subjects design was employed. Participants were randomly assigned to
one of three conditions, Self (n = 26), Similar Race Other (n = 24), or Dissimilar Race
Other (n = 25).
Procedure
The experimental procedure was the same as employed in Pretests 1 and 2.
Participants in all three conditions were provided with a prompt describing the phases of
the experiment and the exercise they would be performing. A research assistant
demonstrated the exercise once to illustrate proper technique and explain how the
82
repetitions would be counted. Then, the participant was fitted in the HMD and immersed
in the virtual environment.
The experiment was structured in three phases that were consistent across
conditions. Participants performed behaviors in physical space while viewing the virtual
room in the HMD. In the first phase, participants performed three sets of 20 exercises
each. In the second phase, participants stood still for two minutes. The third phase was
voluntary: participants were told they could stay in the virtual environment and exercise,
or they could end the experiment.
The phases were the same across conditions, and the stimulus was the same aside
from the manipulation. In every condition, participants viewed a virtual human from the
third person, consistent with the Pretests. That is, the participant did not embody the
virtual human; rather, it was standing in the room facing the participant. The virtual
human started at an average body weight, but participants saw the virtual humans appear
to lose weight as participants physically exercised or gain weight as they remained
inactive. In the first phase, for each repetition of the exercise, the virtual human slimmed
slightly. In the second phase of mandatory inactivity, every three seconds the slimmeddown virtual human regained some weight until it appeared overweight. For the third
phase, the virtual human was reset to the initial weight, and then it appeared to lose or
gain weight in accordance with the participant‘s exercise or inactivity.
The avatars used in this study were designed using the same procedure as in all
the Pretests. Participants had their photographs taken with a digital camera and their
virtual head models were constructed. For the Self conditions, these heads were affixed to
a sex-appropriate generic avatar body. For those assigned to the Similar Race or Different
83
Race conditions, a head of an unknown, same-sex other person was randomly selected
from a pool of previous participants. For those in the Similar Race condition, a head of
the same race as the participant had indicated on their demographic information was
selected. For those in the Dissimilar Race condition, a head of a different race was
selected. For multiracial participants, the head chosen was not of any of the
races/ethnicities that the participant indicated as their own. One update was made in the
avatar building since the Pretests: new, more realistic bodies from the Complete
Characters avatar library were used. Examples of these avatars can be viewed in Figure
12.
Figure 12. Example stimuli for Study 1. Top: a Black female avatar at (from left) normal, slim,
and heavy weight. Bottom: a White male avatar at normal, slim, and heavy weight.
Measures
Items for all measures included in this study can be viewed in Appendix A.
84
Avatar resemblance. Three items were used to assess the self-other manipulation.
Participants were asked to indicate on a 5-point scale (1 = Not at all; 5 = Extremely) the
degree to which they believed the avatar resembled them in the face, the body, and
overall (M = 2.34, SD = .76). Reliability for this measure was Cronbach‘s α = .74.
Avatar race. For manipulation check purposes, participants were asked to identify
the avatar‘s race/ethnicity as White, Black, Latino/a, Asian, or Other.
Racism. To ascertain that the stimulus did not create negative racial associations,
six items from the Modern Racism Scale (McConahay, 1982; McConahay, Hardee, &
Batts, 1981) were adopted. Items were reworded slightly for modernization purposes
(e.g., ―Over the past few years minorities have gotten more economically than they
deserve.‖) The items were embedded with 14 filler items in a purportedly unrelated test
of political attitudes. One item was dropped, leaving the scale consisting of five items
(M = 2.00, SD = .87). Reliability for this measure was Cronbach‘s α = .88.
Identification. Eight items were adapted from Cohen (2001) and used to assess
participants‘ feelings of identification with the avatar. Because the items were originally
written to assess identification with television characters, some were reworded to fit the
stimulus and the term ―my representation‖ was used instead of ―the character.‖
Participants indicated on a 5-point scale (1 = Strongly disagree; 5 = Strongly agree) their
agreement with statements including ―At key moments, I felt I knew exactly what my
representation was going through‖ and ―I think I have a good understanding of my
representation.‖ Responses were averaged; scores ranged from 1.33 to 4.44 (M = 2.84,
SD = .75). A Cronbach‘s alpha of α = .80 was achieved.
85
Presence. The same subjective measure of presence from Pretest 5 was also used.
Eleven items were used to assess participants‘ experience of presence while immersed in
the virtual world. Participants indicated on a 5-point scale (1 = Not at all; 5 = Extremely)
the degree to which they felt present. Responses were averaged; scores ranged from 1.75
to 4.17 (M = 2.89, SD = .66). A Cronbach‘s alpha of α = .90 was achieved.
Specific exercise self-efficacy. Seven items were developed to assess participants‘
self-efficacy towards the specific exercise demonstrated by the virtual model, marching
in place. Bandura‘s (1997) guidelines regarding the level, generality, and strength of
efficacy beliefs were incorporated in the development of the measure. To parallel
exercise performance in the real world, the items were then framed given the Surgeon
General‘s recommendations on exercise (30 minutes a day). Participants indicated on a
five-point scale (1 = Strongly disagree; 5 = Strongly agree) their agreement with
statements including ―I could perform the exercise I saw in the real world at a steady rate
for 30 minutes‖ and ―I would be successful performing the exercise I saw at a steady rate
for 30 minutes.‖ Responses were averaged; scores ranged from 1.00 to 5.00 (M = 3.78,
SD = .99). A Cronbach‘s alpha of α = .92 was achieved.
General exercise self-efficacy. Bandura (1997) noted that an error in self-efficacy
measurement is that it is often aimed too broadly (e.g., at general health behavior rather
than the specific behavior modeled). Since Pretest 3 had indicated that there was an effect
on general exercise behavior after a similar stimulus, however, a measure of general
exercise self-efficacy was added to see if feelings towards exercise in general were
enhanced by the treatment (Sallis, Pinski, Grossman, Patterson, & Nader, 1988). Fifteen
items addressed participants‘ confidence whether they could ―Stick to an exercise
86
program after a long, tiring day at school or work‖ and ―Make time, even on the
weekends, to exercise‖ (1 = Very confident I could not do it; 5 = Very confident I could
do it). Responses were averaged; scores ranged from 1.20 to 5.00 (M = 3.60, SD = .91). A
Cronbach‘s alpha of α = .95 was achieved.
Exercise repetitions. As in Pretests 1 and 2, the behavioral variable of interest was
exercise repetitions. A research assistant counted each exercise repetition participants
performed during the voluntary third phase of the experiment and recorded each with a
keystroke. Fifty-seven participants (76%) chose to exercise during the third phase; the
number of repetitions performed ranged from 0 to 189 (M = 40.39, SD = 40.80).
Results
Manipulation Checks
Participants were asked to note the degree to which they felt the avatar resembled
them to determine whether the self-other manipulation was successful. There was a
significant difference across conditions, F(2, 72) = 19.37, p < .0005, partial η2 = .35.
Follow-up tests with Bonferroni corrections indicated that participants in the Self
condition (M = 2.76, SD = .66) rated the avatars as resembling them more than those in
the Dissimilar condition (M = 1.72, SD = .58), t(49) = 5.95, p < .0005, Cohen‘s d = 1.67,
and those in the Similar condition (M = 2.53, SD = .62) perceived more resemblance than
those in the Dissimilar condition, t(47) = 4.73, p < .0005, Cohen‘s d = 1.35. However,
there was no significant difference between the Self and Similar conditions, t(48) = 1.26,
p > .05, Cohen‘s d = .36. Thus, the Self manipulation failed.
Participants were also asked to indicate the avatar‘s race. Fifty-one participants
(68%) correctly identified the avatar‘s race and 24 (32%) did not. Of those that did not, 6
87
(8%) were in the Dissimilar condition and although they did not identify the avatar‘s race
correctly, did identify the race as one dissimilar to their own. Thus, a total of 57
participants (74%) correctly recognized that the avatar‘s race was similar or dissimilar to
their own and the manipulation failed for 18 participants (26%). Surprisingly, this failure
to distinguish the avatar's race was not isolated to the Other conditions, as 26.9% of
participants in the Self condition (7 of 26) failed to correctly identify the avatar's race.
Indeed, a chi-square analysis revealed there was no significant differences across
conditions on correct identification of the avatar's race, χ2 [N = 75] = .64, p > .05.
Finally, participants‘ attitudes about race were assessed. There were no significant
differences across conditions, F(2, 65) = .79, p > .05, partial η2 = .02. Participants in the
Self (M = 1.90, SD = .93), Similar (M = 1.91, SD = .82), and Dissimilar (M = 2.19,
SD = .84) conditions did not differ on the Modern Racism Scale.
Hypotheses and Research Questions
Means and standard deviations for all dependent variables can be viewed in Table
3.
88
Table 3
Study 1 Means and Standard Deviations for Dependent Variables by Condition and Sex
Exercise
Repetitions
Resemblance
Identification
Specific SE
M
2.76
SD
.66
M
2.88
SD
.83
M
3.76
SD
1.08
M
41.54
SD
40.67
M
2.87
SD
.72
M
3.81
SD
.96
Men
2.95
.62
2.93
.82
3.48
1.22
51.46
53.81
2.99
.75
4.04
.71
Women
2.56
.67
2.84
.86
4.04
.89
31.62
19.70
2.76
.69
3.59
1.15
2.53
.62
2.71
.68
3.96
.90
47.58
43.53
2.90
.63
3.51
1.04
Men
2.58
.59
2.48
.72
4.25
.87
45.41
47.82
2.88
.71
3.91
.90
Women
2.47
.67
2.94
.58
3.68
.87
49.75
40.80
2.92
.55
3.11
1.05
1.72
.58
2.92
.74
3.63
.98
32.28
38.03
2.91
.67
3.46
.70
Men
1.61
.57
3.15
.80
3.89
.97
41.41
40.93
3.21
.63
3.73
.46
Women
1.82
.59
2.70
.65
3.39
.96
23.85
34.60
2.63
.59
3.22
.80
2.34
.76
2.84
.75
3.78
.99
40.39
40.80
2.89
.66
3.60
.91
DV
Condition
Self
Similar
Dissimilar
Overall
Presence
General SE
Identification. To test H1, a one-way ANOVA was performed on identification.
No significant difference in identification was found between conditions, F(2, 72) = .51,
p > .05, partial η2 = .01. H1 was not supported.
Specific self-efficacy. For H2, a one-way ANOVA was performed on specific
exercise self-efficacy. No significant difference in specific self-efficacy was found
between conditions, F(2, 72) = .70, p > .05, partial η2 = .02. H2 was not supported.
Exercise repetitions. For H3, a one-way ANOVA was performed on exercise
repetitions. No significant difference in exercise repetitions was found between
conditions, F(2, 72) = .51, p > .05, partial η2 = .01. H3 was not supported.
89
General self-efficacy. For RQ1, a one-way ANOVA was performed on general
exercise self-efficacy. No significant difference in general self-efficacy was found
between conditions, F(2, 72) = 1.10, p > .05, partial η2 = .03.
Presence. For RQ2, a one-way ANOVA was performed on presence. No
significant difference in presence was found between conditions, F(2, 72) = .02, p > .05,
partial η2 = .00.5
Additional Analyses
Because the manipulations themselves failed, further analyses were run. First, sex
was considered as a factor. Then, the conditions were variously combined to examine
whether previous work was conceptually supported. Finally, the stimuli were reexamined and recoded on variables that may have influenced the findings.
Although sex had not been a significant factor in any of the Pretests involving
exercise, sex was entered as a factor and 3 by 2 factorial ANOVAs were run on the
variables of interest (identification, specific exercise self-efficacy, exercise repetitions,
presence, and general self-efficacy). Results of these analyses can be viewed in Appendix
B. The only significant effect was a main effect on general self-efficacy, indicating that
men reported higher general exercise self-efficacy than women, F(1, 69) = 8.29, p < .01,
partial η2 = .11. One other effect bordered on significance, an interaction between
condition and sex on specific self efficacy, F(2, 68) = 2.68, p = .076, partial η2 = .07.
There was a trend wherein men reported more specific self-efficacy in the Similar Race
condition than the Dissimilar Race condition and the Self condition. The trend for women
followed the hypothetical relationship; they reported more specific self-efficacy in the
90
Self condition than the Similar Race and Dissimilar Race conditions, as can be seen in
Figure 13.
Figure 13. Specific self-efficacy by Condition and Sex. The interaction effect reached borderline
significance (p = .076).
Combining the assigned conditions in analyses also did not support hypothetical
relationships. In light of previous studies, we would expect that the Self would elicit more
exercise than the Other categories, but this was not supported by the data; there were no
significant differences between the means of the Self (M = 41.54, SD = 40.97) and
combined Similar Race Other and Dissimilar Race Other (M = 39.83, SD = 41.55)
conditions, t(72) = .17, p > .05. There were also no differences in identification,
presence, specific exercise self-efficacy, or general exercise self-efficacy using these
91
distinctions; means, standard deviations, and results of these analyses can be viewed in
Appendix C. Alternatively, the Self and Similar Race Other categories, given they were
determined not to be considered different in resemblance, could be combined and
contrasted to the Dissimilar Race condition. However, independent samples t-tests
revealed no differences between these two categories either. Means, standard deviations,
and results of these tests can be viewed in Appendix D. Clearly, the conditions designated
by the experimenter failed in some fundamental way.
Given these unexpected findings, as well as participants‘ spontaneous responses
and the experimenters‘ observations, further analyses on the stimuli themselves were also
pursued. Two independent coders (one male, one female) were recruited to code all of the
heads used as stimuli on attractiveness and naturalness (i.e., whether the head looked
natural or computer-generated). For the Self heads, the coders also measured the
attractiveness of the photos the heads were based on (to determine if there was a
discrepancy between the participant‘s attractiveness and their head‘s attractiveness) as
well as the degree to which the head accurately represented the photograph. Following
Hayes and Krippendorff (2007), Krippendorff‘s alpha for interval-level measurements
was used to calculate reliability, which was deemed acceptable at α = .74. Coders‘ ratings
were then averaged within the coding categories.
The coding revealed that there was a significant difference in between the
participants‘ attractiveness in photos (M = 3.17, SD = .55) and the attractiveness of their
heads (M = 2.44, SD = .68), t(25) = 4.38, p < .0005. There were also significant
differences in head attractiveness across conditions, F(2, 72) = 14.49, p < .0005, partial
η2 = .29. A follow-up LSD test revealed that heads from the Similar (M = 3.29, SD = .75)
92
and Dissimilar (M = 3.44, SD = .71) conditions were ranked as more attractive than those
in the Self condition (M = 2.44, SD = .68). The heads also differed significantly on
naturalness, F(2, 72) = 3.94, p < .05, partial η2 = .10. A follow-up LSD test revealed that
heads from the Similar (M = 3.79, SD = 1.03) and Dissimilar (M = 3.70, SD = .74)
conditions were more natural-looking than those in the Self condition (M = 3.13,
SD = .92).
Because attractiveness has been shown to be an important factor in model
imitation (i.e., observers are less likely to imitate unattractive models), participants whose
models had heads that were rated below 3 (the midpoint of the scale) were eliminated,
and the remainder were analyzed. This process also eliminated heads that were ranked as
below average on naturalness. Previous tests had shown the discrepancy in attractiveness
across conditions, and when the unattractive heads were removed, the Self condition
(n = 10) was disproportionately impacted compared to the Similar (n = 21) and
Dissimilar (n = 21) conditions. Also, limiting the sample size adversely impacted the
power to detect differences (Cohen, 1988). Thus, the following analyses should be
considered cautiously.
A one-way ANOVA was run to see if the predicted effect on exercise modeling
occurred; the test showed a borderline significant difference, F(2, 49) = 5.48, p = .075,
partial η2 = .29. The means were in the predicted direction. Followup LSD tests revealed
that although there were no differences between Self (M = 64.80, SD = 53.09) and
Similar Other (M = 48.05, SD = 39.85), Self was significantly different than Dissimilar
Other (M = 29.48, SD = 34.86); see Figure 14. No difference was found between
participants in the Similar and Dissimilar conditions.
93
Figure 14. Effect of Condition on exercise repetitions after removal of unattractive stimuli. The
effect was borderline significant (p = .075).
Regarding specific self-efficacy, although the means were also in the predicted
direction, a one-way ANOVA revealed no differences between Self (M = 4.21, SD = .77),
Similar Other (M = 3.95, SD = .92), and Dissimilar Other (M = 3.58, SD = 1.02),
F(2, 49) = 1.76, p = .18, partial η2 = .07. Thus, eliminating problematic subjects yielded
means that generally followed the trends anticipated by the hypotheses for exercise
repetitions and specific self-efficacy. No significant differences were found for
identification [F(2, 49) = 1.33, p > .05, partial η2 = .05], presence [F(2, 49) = .43, p > .05,
partial η2 = .02], or general self-efficacy [F(2, 49) = .67, p > .05, partial η2 = .03].
Additionally, further analyses were conducted to examine participants‘ responses
independent of the assigned conditions. Pearson correlations were run to see if other
94
relationships existed among the dependent variables. These tests indicated that
resemblance, identification, presence, and exercise repetitions were all significantly and
positively correlated, although none of these were related to specific or general selfefficacy. These correlations can be examined in Table 4.
Table 4
Pearson Correlations between Study 1 Dependent Variables (N = 75)
Spec.
DV
Resemblance Identification
Gen.
Presence
Exercises
SE
SE
Resemblance
--
.31**
.29*
.12
.02
.37**
Identification
.31**
--
.77**
-.03
-.06
.45**
Presence
.29*
.77**
---
-.01
-.22
.36**
Specific SE
.12
-.03
-.01
--
.16
.09
General SE
.02
-.06
-.22
.16
--
-.13
.37**
.45**
.36**
.09
-.13
--
Exercises
**Correlation is significant at p < .01. *Correlation is significant at p < .05.
The assertions of this line of research are that the more a person feels an avatar
resembles the self and the more he or she identifies with the avatar, the more inclined the
individual will be to imitate an avatar‘s actions. Based on these assertions and the
observed correlations, a model was hypothesized. A regression was run using exercise
repetitions as the dependent variable and resemblance, identification, and presence as
predictor variables. Regression analysis indicated that the set of predictors performed
better than chance at predicting exercise repetitions, R = .51, Adjusted R2 = .23,
F(3, 72) = 8.28, p < .0005. Two of three predictors significantly predicted exercise
95
repetitions. Higher levels of perceived resemblance were associated with more exercise
repetitions, β = .25, t = 2.35, p < .05. Higher levels of identification were also associated
with more exercise repetitions, β = .36, t = 2.22, p < .05. Presence was not a significant
predictor, β = .02, t = .10, p > .05.
Discussion
The major issue with this study is that the intended manipulations failed.
Participants did not perceive the Self avatars as resembling them more than the Similar
avatars. Thus, it is unsurprising that we did not see a replication of previous results
wherein Self avatars yielded more exercise behaviors; in previous studies the participants
always perceived the Self avatars as resembling the self more than Other avatars, which
is believed to be the mechanism that motivates participants to model the behavior. It
should also be noted that the race manipulation was also unsuccessful for one quarter of
participants. Due to the failure of the conditions in creating clearly delineated categories
based on degrees of self-resemblance as defined by race, the data were re-examined
based on the fundamental variables believed to influence outcomes.
Another issue with the conditions is that there were unintended qualitative
differences. The Self avatars were determined to be less attractive and less natural than
the avatars in the Similar and Dissimilar conditions. Additionally, some of the Self
avatars were determined to be inaccurate representations of the participants. These
discrepancies undoubtedly influenced the findings. People react more positively to
attractive people and are more persuaded by them as opposed to unattractive people
(Dion, Berscheid, & Walster, 1972; Langlois, Kalakanis, Rubenstein, Larson, Hallam, &
Smoot, 2007; Reihnard, Messner, & Sporer, 2006). Observers are also more likely to
96
imitate attractive models than unattractive ones (Bandura, 1977; Van Leeuwen, Veling,
Van Baaren, & Dijksterhuis, 2009). Yee and Bailenson (2007) also found that people
who embodied unattractive avatars disclosed less and maintained more personal distance
to a confederate than those in attractive avatars. Thus, viewing an unattractive head,
particularly one that is discrepant from one‘s own attractiveness, may have influenced the
findings, particularly in the Self condition because more than half of the heads were
determined to be unattractive or unnatural.
Removing the problematic heads from the sample and reanalyzing the remaining
participants revealed a trend toward supporting previous findings that the virtual self is
more effective than other representations in encouraging modeling behavior. The means
also gave a general indication that the degree of similarity was related to the number of
exercises performed, with Self avatars yielding the most exercise, followed by Similar
avatars, and Dissimilar avatars promoted the least. Future studies should consider the role
of attractiveness as well as similarity when selecting models.
In line with previous findings and conceptual predictions, the more participants
felt the avatar resembled them and the more they identified with the avatar, the more
exercises they performed. These findings, in addition to the Pretests, support the
arguments of social cognitive theory and the proposed role of identification in behavioral
modeling. Although the manipulated conditions were not effective in creating the
variation in resemblance and identification that they intended to, the findings still support
the conceptual assertions of this research: when people observe an avatar, the more they
believe it resembles them, the more they identify with it, and the more likely they are to
imitate the avatar‘s modeled behavior. Thus, more careful construction of these virtual
97
representations of the self to maximize resemblance should continue to prove effective at
motivating people to exercise.
It is interesting to note that only some borderline effects were found on exercisespecific self-efficacy despite the differences in identification and, in some analyses,
exercise performance. One possibility is that the measure was flawed. Although the
measure was adapted from suggestions made by Bandura (1997) and it proved to be
reliable, it has not been validated elsewhere. The measure did not perfectly parallel the
stimulus; the items considered a 30 minute time frame to reflect effective real world
exercise conditions, whereas the stimulus did not last 30 minutes. Future studies should
ensure a direct parallel to the stimulus (indeed, a correction that informed the scale
development in Study 2). Another possibility supported by the model above is that the
identification evoked by a self-resembling avatar independently influences modeling
without impacting the individual‘s cognitions regarding his or her ability to perform the
exercise. Future studies should continue to measure exercise-specific self-efficacy to see
if it can be influenced by virtual self models, and the construct should also be examined
alongside identification to determine their independent contributions to the variance in
modeling behavior.
One explanation for the failure of the Self manipulation is that most of the heads
made for this condition were made specifically for this study, whereas the Other heads
were drawn from previous participants. In the past, considerably more time and technical
personnel were allotted towards the crafting of the heads, allowing for more time to use
Photoshop to clean up poor photographs (e.g., those with uneven lighting) and more time
to test and rebuild heads that were not optimal. Unfortunately, limited resources and
98
deadlines did not allow for as much processing with new photographs and participants,
and this inadvertently created a difference in the quality between the Other heads and the
Self heads. People may have reacted negatively to an inaccurate and less attractive
version of themselves. The experience may have been aversive enough to make
participants not want to interact with this version of the self, hence limiting their desire to
exercise and remain in the environment wherein they would continue to see the
misrepresentation face-to-face. In the future, head quality should be maximized and
pretested to ensure qualitative differences do not exist between conditions.
Study 1 attempted to manipulate stimuli in order to impact participants‘ exercise
behaviors and specific self-efficacy. It did not consider the role that pre-existing selfefficacy may play in how observers respond to virtual selves. Study 2 was designed to
consider its role and also to compare virtual environments with a new form of treatment,
mental imagery.
99
CHAPTER ELEVEN: DISSERTATION STUDY TWO: COMPARING VIRTUAL
AND IMAGINED STIMULI FOR HIGH AND LOW SELF-EFFICACY
PARTICIPANTS
Another lingering question from this line of research is a practical one: is an
immersive virtual environment necessary to evoke the exercise behavior response?
Immersive virtual environments are costly and not widely available; creating the virtual
stimulus, particularly building the virtual self, is time-consuming; and the equipment can
be considered cumbersome. If the same effects can be yielded using mental imagery, a
cost-free and unencumbered method with minimal time invested in stimulus creation,
then many of the supporting arguments for the advantages of treatments implementing
immersive virtual environments become null. Thus, Study 2 was designed to compare the
effects of virtual reality and mental imagery stimuli on low and high self-efficacy
participants.
Mental Imagery
Mental imagery (MI) is defined as a cognitive activity in which one perceives a
stimulus without the stimulus being present, possibly leading to a subjective experience
as if the stimulus actually were present (Kosslyn, 1981; Kosslyn, Thompson, & Ganis,
2006). Mental practice (MP), more specifically, is the ―symbolic, covert, mental rehearsal
of a task in the absence of actual, overt, physical rehearsal‖ (Driskell, Copper, & Moran,
1994, p. 481). Mental imagery has been shown to be a powerful motivational tool that
can alter subsequent attitudes and behaviors. Typically, mental imagery is invoked using
a written prompt or oral instructions that give the audience specific instructions on what
they should be sensing in their ―mind‘s eye.‖ Participants close their eyes and use their
100
cognitive resources to recreate a scene that may focus on the details of the performance
of a particular behavior, the environment in which it takes place, or the outcomes that it
may yield. The use of mental practice has been shown to promote task self-efficacy as
well as enhanced skill performance (Cumming, 2008; Driskell et al., 1994; Feltz, 1984;
Martin & Hall, 1995). Visualization techniques are commonly used to promote healthy
behaviors as well as to develop athletic skills (Hausenblas, Hall, Rodgers, & Munroe,
1999). Mental practice of exercise can even lead to the same physiological responses as
actual exercise (Wang & Morgan, 1992).
Like virtual environments, mental imagery provides the opportunity to create
stimuli that may not be possible in the physical world. Mental imagery also allows for the
individual to create a rich, multi-channel stimulus similar to what can be created using
virtual technologies. However, virtual environments do have some advantages to
visualization techniques. Some people may have trouble imagining a particular scene
because they do not have sufficient knowledge; for example, a runner might not know
how her legs should be positioned to effectively clear a hurdle. A virtual environment
may be seen as more entertaining or interesting than a visualization task (Vorderer,
2000), so participants may be more motivated or engaged in a virtual environment.
Additionally, there is the question of required mental effort.
Cognitive Load
Miller (1956) was among the first psychologists to acknowledge the limits of our
working memory, suggesting that there is only so much information we can attend to as
we actively engage in cognitive processing. According to Chandler and Sweller‘s (1991)
cognitive load theory, any channel is limited in the amount of cognitive processing that
101
occurs at any given time. If we are observing something through a visual channel, for
example, our resources are divided among interpreting physical representations (i.e., what
we see in front of us) and accessing related memory representations, such as our previous
experience with those representations (Mayer & Moreno, 2003). When we must generate
a stimulus, such as through visualization, we must simultaneously access our deep
working memory representations (e.g., visual models) to facilitate the creation of those
images, access our related memory representations, and also interpret the representations
that we are generating (Mayer & Moreno, 2003). Thus, having the cognitive
responsibility of simultaneously generating these images may result in cognitive
overload, or the point at which needed cognitive processing exceeds the individual‘s
capacity to process (Chandler & Sweller, 1991).
Thus, one major drawback to mental imagery and mental practice is the amount of
cognitive effort required. Participants must actively combat other intervening thoughts,
remain focused on the imagery task, and effectively render it in their minds. Visualization
thus requires a great deal of cognitive effort, which may explain why it is not always
successful (Kosslyn et al., 2006). Ginns, Chandler, and Sweller (2003) found that when
students had to imagine novel scenarios, it was more effective for them to absorb that
knowledge visually; students who imagined the scenarios learned significantly less than
those who learned through a visual channel, likely because of the cognitive effort exerted
in trying to visualize. Babin and Burns (1998) suggest that part of this cognitive load may
come from imagery elaboration, wherein effort is expended to associate the information
in the mental stimulus with other information in long-term memory. Because the imager
is simultaneously trying to render the image and remain engaged with it while
102
summoning the schematic information associated with the rendered image, cognitive
overload may occur. Participants may resort to solely focusing on the image, shutting
down elaboration, or elaborate, which takes away cognitive resources from the imaging
effort. Mediated stimuli do not require such a tradeoff because participants to not have to
actively render those images; however, features of mediated messages may still create
cognitive overload (Lang, 2000).
The limited-capacity information-processing model (Lang, 1995, 2000) suggests
that media observers are equipped with limited cognitive resources to handle mediated
messages, and that media messages are comprised with many elements that elicit
automatic cognitive responses from observers. Several factors contribute to the overall
cognitive load of a message or image, including the number of features in the message as
well as the amount of information that is not provided in the message or implied (i.e.,
what must be filled in by the receiver; Kosslyn et al., 2006; Lang, 1995, 2000). The
model argues that different media elicit different responses due to different channels
(e.g., radio is restricted to audio whereas television features audio and visual information)
and features, and tests of the model have been conducted across radio (Lang, Schwartz,
Lee, & Angelini, 2007; Potter, 2000), television (Geiger & Reeves, 1993; Lang, Geiger,
Strickwerda, & Sumner, 1993; Lang, Newhagen, & Reeves, 1996; Lang, Bolls, Potter, &
Kawahara, 1999) and the Internet (Lang, Borse, Wise, & David, 2002).
No work was identified that has tested the model within virtual environments;
however, logical conclusions may be drawn on the similarities of the stimuli within the
VE. These studies demonstrate that cognitive overload tends to happen when a stimulus
requires maximal attention or cognitive effort to process. Although many stimulus
103
features, including movement, novelty, or change, yield an immediate orienting response
upon their introduction, this arousal wanes as the person adjusts to the stimulus, usually
within three to seven seconds (Reeves, Lang, Kim, & Tatar, 1999; Reeves & Nass, 1996).
Once participants have acclimated to the message, barring novel or emotional content,
participants experience less arousal and exert less cognitive effort (Lang, 2000; Lang et
al., 2007). The stimuli used in these studies have been limited to the visual channel; no
other sensory channels are incorporated. Thus, while there is potential for cognitive
overload during the orientation response to the stimulus within the virtual environment, it
is anticipated that after the observer has acclimated, no additional cognitive strain would
occur if the stimulus remains steady. Thus, participants can attend to the stimulus and the
message can be fully processed, achieving maximal effects.
In contrast, a visualization task requires constant cognitive effort, perhaps to the
point of overload. The images must be generated, maintained, and attended to
simultaneously. Having to devote so many cognitive resources to this process may
hamper learning and thus limit the effects of the message, as Ginns et al. (2003) found.
Research also indicates that the ability to create mental imagery varies among individuals
(Hall & Martin, 1997), so these tasks may not be effective treatments for those who have
difficulty visualizing. In contrast to the effortful nature of mental practice, a virtual
stimulus requires less cognitive labor. Thus, virtual environments may have an advantage
over the use of mental practice to promote healthy behavior, especially among those
individuals who have difficulty conjuring mental images on demand. Currently, however,
no studies have addressed the comparison of the two stimuli in the health domain.
104
Thus, we would anticipate differences between a visual virtual environment, in
which the scene is rendered before the individual‘s eyes, and a mentally-imaged
environment that the individual must generate. To respond to these unanswered
questions, an exploratory study comparing the two stimuli was designed. Given that no
previous research has compared these two stimulus types in the domain of behavior
modeling, the study was framed to gather as much information as possible on potential
differences. Additionally, this study sought to extend the work of Study 1 and, in this
case, examine participants pre-existing self-efficacy regarding an exercise behavior
before the administration of any treatment, thus enabling a comparison on the potential
impact on both self-efficacy and exercise. The following hypotheses and research
questions are proposed:
H4: Participants who receive a VE stimulus will show a greater increase in their
estimated pushup ability than participants who receive an MP stimulus.
RQ3: Will pre-existing self-efficacy interact with stimulus type to affect changes
in estimated pushup ability?
H5: Participants who receive a VE stimulus will demonstrate higher levels of
specific self-efficacy than those who receive an MP stimulus.
RQ4: Will pre-existing self-efficacy interact with stimulus type to affect specific
self-efficacy?
H6: Participants who receive a VE stimulus will show a greater increase in actual
pushup performance than participants who receive an MP stimulus.
RQ5: Will pre-existing self-efficacy interact with stimulus type to affect changes
in actual pushup performance?
105
No relevant research was identified regarding the comparison of psychological
effects of encountering the self within imagined and virtual environments. Imagining
oneself, which is a common occurrence when we recall memories or daydream, and
encountering a three-dimensional virtual representation of oneself, which is likely a rare
if not novel event, are clearly two distinct experiences. Thus, no specific predictions were
made and exploratory research questions were proposed regarding perceived
resemblance, identification, and presence.
RQ6: Will there be a difference in perceived resemblance across conditions?
RQ7: Will there be a difference in identification across conditions?
RQ8: Will there be a difference in presence across conditions?
Method
Sample
A convenience sample was obtained from the student population of a mediumsized West Coast university. The sample (N = 91) consisted of 57 women and 34 men
aged 18 to 53 (M = 20.89, SD = 5.15). Participants self-reported their race/ethnicity as
Caucasian/European-American/White (45.1%; n = 41); African/African-American/Black
(9.9%; n = 9); Latino/a (16.5%; n = 15); Asian/Asian-American (18.7%; n = 17); Pacific
Islander (1.4%; n = 1); and Other or multiple races/ethnicities (8.8%; n = 8).
Design
A between-subjects design was employed for this experiment. Participants were
randomly assigned to either a Virtual Environment (VE) or Mental Practice (MP)
stimulus. Participants‘ responses on the pretest determined whether they were categorized
as low or high self-efficacy. If participants indicated they ―probably‖ or ―definitely‖
106
could not perform the number of pushups in the stimulus (30 for women and 60 for men),
they were categorized as low self-efficacy; all others were categorized as high selfefficacy. Participants were randomly assigned to either the VR or MP treatment. Thus,
there were four conditions for analysis: VE-High SE (n = 19), VR-Low SE (n = 24), MPHigh SE (n = 17), and MP-Low SE (n = 31).
Procedure
To determine the number of pushups for the stimulus, a pilot survey was
conducted using a class pool (N = 66; 20 men and 46 women). Participants were asked to
time themselves completing as many pushups as possible in one minute, and then respond
to ten items on a 5-point scale (1 = Definitely could not do it; 5 = Definitely could do it)
whether or not they felt they could complete a certain number of pushups in one minute
(5, 10, 15, 20, 25, 30, 40, 50, 60, more than 60). Scatterplots of these responses were then
examined to determine a clear and appropriate cut-off point for each of the sexes. It was
determined that 30 pushups for women and 60 pushups for men created a clear split for
each group; that is, nearly half of the participants indicated they ―definitely‖ or
―probably‖ could not do it, and nearly half indicated they ―definitely‖ or ―probably‖
could.
Participants in this study appeared for a photo session wherein they had their
photographs taken with a digital camera and later their virtual head models were
constructed as they had been in all studies. Approximately one week before their
experiment was scheduled, participants were sent an email with a link to a pretest. As in
the pilot survey, the experimental pretest asked them to rank their confidence in
107
performing specific numbers of pushups in one minute. Participants also completed an
assessment of their mental imagery ability at that time.
In the mental practice condition, participants were asked to stand in the
experimental room. Low self-efficacy participants were given the following instructions:
For this task, you‘re going to be imagining a room that resembles this lab. Inside
that room, you will imagine your physical self. You will be looking at yourself
from the 3rd person, as if looking at a photograph of yourself. The person you will
imagine in the room is a mental image of you.
While imagining this world, you are going to envision your self performing an
exercise, pushups. Your self will be performing 30 pushups [for women; 60 for
men] in one minute. When we asked you earlier, you indicated that you are unable
to perform 30 pushups [for women; 60 for men] in one minute. Your task will be
to imagine yourself successfully perform pushups at a rate that you cannot
perform in the real world.
The instructions for high self-efficacy participants were the same, except the last
sentences read:
When we asked you earlier, you indicated that you are able to perform 30 pushups
[for women; 60 for men] in one minute. Your task will be to imagine yourself
successfully perform pushups at a rate that you can perform in the real world.
Next, participants were instructed to look around the room for 30 seconds to familiarize
themselves with the setting. The experimenter timed this phase and let them know when
30 seconds had transpired. Then, participants were instructed to close their eyes and
spend 60 seconds simply imagining themselves standing in the room from the third
person. The experimenter kept time and told the participants when to stop. In the final
phase, participants were instructed to close their eyes and spend 60 seconds imagining
themselves successfully completing 30 (for women) or 60 (for men) pushups. Women
were told this was approximately one pushup every two seconds; men were told this was
108
one pushup each second. Afterward, participants were instructed to complete some
measures on the computer.
In the virtual environment condition, participants stood at the same point in the
experimental room as the MP participants. They were given the following instructions:
For this task, you‘re going to be put in a room that resembles the actual lab. Inside
that room, you are going to see a virtual human. This isn‘t a real person—it is a
virtual person we have created in the virtual lab. This virtual human was built to
look like you. We took the photographs you took earlier and used them to build
this virtual representation of you. The person you will see in the room is a virtual
copy of you.
While inside the virtual world, you are going to see your avatar performing an
exercise, pushups. Your avatar will be performing 30 [for women; 60 for men]
pushups in one minute. When we asked you earlier, you indicated that you are
unable to perform 30 [for women; 60 for men] pushups in one minute. Your task
while in the virtual world will be to watch your avatar successfully perform
pushups at a rate that you cannot perform in the real world.
Again, high self-efficacy participants were given the same prompt, except that the last
two sentences read:
When we asked you earlier, you indicated that you are able to perform 30 pushups
[for women; 60 for men] in one minute. Your task while in the virtual world will
be to watch your avatar successfully perform pushups at a rate that you can
perform in the real world.
Then, participants were outfitted in the head-mounted display. The room was rendered in
front of them from the same perspective. Next, they were instructed to look around the
room for 30 seconds to familiarize themselves with the setting. The experimenter timed
this phase and let them know when 30 seconds had transpired. Then, participants were
timed as they spent 60 seconds looking at their virtual self-representation standing in the
room across from them (i.e., in the third person). The experimenter kept time and told the
participants when to stop. In the final phase, participants watched as the virtual self
successfully completed 30 pushups (women) or 60 pushups (men) in 60 seconds. See
109
Figure 15 for examples of the virtual stimuli. Afterwards, participants were taken out of
the equipment and instructed to fill out measures on the computer.
Figure 15. Example stimuli from Study 2. Participants observed their virtual selves
standing for one minute (left) before observing them perform pushups successfully
(right). Numbers to the right of the virtual self counted off the number of pushups as a
stopwatch (at top) kept time.
At this point, both groups completed measures indicating their level of selfefficacy regarding the pushup exercise. Once they completed this part of the survey, the
exercise phase commenced. The experimenter verbally explained proper pushup
technique and then demonstrated once for the participant. After making sure the
110
participant understood proper form, the experimenter led the participant back to the
experimental room. Participants were then instructed:
I am going to ask you to perform as many pushups as you can in one minute while
maintaining proper form. If you wish, you can pause and rest in between
exercises. If you experience any pain aside from regular muscle exertion, please
let me know and do not continue. The goal is to perform as many pushups as
possible in one minute while maintaining good form. If you start to lose your
form and do not complete the exercise properly, I will let you know; those
repetitions will not count. I will count out loud so you know how many pushups
you have completed, and help you keep track of time by telling you when you‘ve
reached 30 seconds, the halfway point.
The experimenter asked the participant to get in position and then started the timer as the
participant began. Only pushups completed with proper form were counted; if
participants‘ form was not acceptable, the experimenter offered corrective feedback (e.g.,
―You have to bend your elbows and go further down‖) and did not count that repetition.
After one minute, the experimenter informed the participant that the exercise phase was
complete, and participants returned to the computer to complete the rest of the survey
measures.
Pretest Measures
Time 1 pushups. Participants were instructed on proper pushup form. Next, they
were asked to time themselves for one minute and complete as many pushups as possible.
Participants reported completing between 0 and 63 pushups (M = 23.30, SD = 15.50).
This item was self-reported.
Time 1 pushup self-efficacy. After completing the pushup task, participants were
asked to respond to twelve items to determine their confidence levels in performing a
certain number of pushups (ranging from 5 to more than 80) in one minute. Participants
indicated on a 5-point scale (1 = Definitely could not do it; 5 = Definitely could do it)
111
whether they felt they could perform that number of pushups in one minute. This data
was used to divide participants into low and high self-efficacy groups.
Experiment Measures
Time 2 estimated pushups. Upon arrival for the experiment, participants were
asked to estimate how many pushups they felt they could complete in one minute
(M = 19.86, SD = 13.63).
Resemblance. Participants were asked to indicate on a 5-point scale (1 = Not at
all; 5 = Extremely) the degree to which they believed the representation (avatar in VR
conditions or imagined self in MI conditions) resembled them in the face, the body, and
overall (M = 3.09, SD = 1.01). Reliability for this measure was Cronbach‘s α = .85.
Identification. The same items as Study 1 were used to assess identification.
Scores ranged from 1.00 to 5.00 (M = 3.05, SD = .91). A Cronbach‘s alpha of α = .85 was
achieved.
Presence. The same measure as Study 1 was used to assess presence. Scores
ranged from 1.00 to 5.00 (M = 2.72; SD = .68). A Cronbach‘s alpha of α = .86 was
achieved.
Specific exercise self-efficacy. The same items used in Study 1 were adapted to
assess participants‘ self-efficacy towards the specific exercise demonstrated by the virtual
model, pushups. Participants indicated on a five-point scale (1 = Strongly disagree;
5 = Strongly agree) their agreement with statements including ―I would be successful in
performing pushups at the pace I saw‖ and ―I would have no problems performing
pushups at the pace I saw.‖ Responses were averaged; scores ranged from 1.00 to 4.43
(M = 2.22; SD = .93). A Cronbach‘s alpha of α = .94 was achieved.
112
Time 3 estimated pushups. After exposure to the treatment but before they
completed any exercises, participants were asked again how many pushups they believed
they could complete in sixty seconds. This estimate was taken to compare to the Time 2
estimate that participants provided before the treatment.
Time 3 pushups. At the time of the experiment, a research assistant counted each
pushup participants performed with proper form in sixty seconds during the exercise
phase. The number of performed ranged from 0 to 67 (M = 27.62; SD = 16.44).
Results
Because of a) the difference in sample sizes and individual cell sizes between
men and women; b) the difference in stimuli; and c) the difference in men‘s and women‘s
upper body strength and the ability to perform pushups, men and women were analyzed
separately. Factorial ANOVAs were conducted to examine the effects of stimulus and
pre-existing self-efficacy for each sex. Levene‘s tests were run to test for equality in
variance given the inequal cell sizes; variances were equal unless otherwise noted.
The means and standard deviations for all dependent variables can be seen in
Table 5.
113
Table 5
Study 2 Means and Standard Deviations for Dependent Variables by Condition and Sex
DV
T1 Pushups
T2 Est. Pushups
Specific SE
T3 Est. Pushups
T3 Pushups
Condition
M
SD
M
SD
M
SD
M
SD
M
SD
VE Low
24.31
15.29
20.18
14.35
2.14
.70
25.56
17.27
26.63
16.10
Men
34.00
7.45
31.50
6.99
2.39
.80
36.30
11.00
37.77
12.15
Women
8.17
10.05
6.60
6.72
1.84
.43
7.67
7.89
13.46
8.15
35.39
16.75
32.65
14.59
3.74
1.10
36.39
17.89
42.21
16.23
Men
48.50
14.54
44.29
15.39
3.77
1.06
50.83
16.56
56.75
14.60
Women
24.14
8.09
24.50
6.43
3.71
1.19
24.00
4.87
31.64
5.71
17.13
13.51
15.47
12.07
1.82
.88
15.80
12.28
23.07
15.36
Men
34.33
6.63
31.44
5.20
1.92
.80
31.56
6.29
43.44
7.07
Women
9.76
7.53
8.62
6.01
1.78
.93
9.05
6.58
14.73
8.21
37.33
16.19
30.47
15.23
3.45
.89
33.42
16.00
35.77
13.11
Men
57.00
8.71
45.00
19.15
3.50
.92
50.50
12.15
52.75
14.08
Women
27.50
6.91
25.18
9.97
3.43
.93
24.88
9.42
30.54
7.34
25.51
16.95
22.86
15.33
2.61
1.20
24.75
17.15
30.37
16.89
VE High
MP Low
MP High
Overall
114
Table 5 (continued)
Study 2 Means and Standard Deviations for Dependent Variables by Condition and Sex
DV
Resemblance
Identification
Condition
M
SD
M
SD
M
SD
VE Low
2.31
.77
2.34
.82
2.35
.62
Men
2.49
.57
2.48
.82
2.57
.62
Women
2.09
.93
2.18
.83
2.09
.54
2.74
.59
2.80
.52
2.58
.44
Men
2.50
.71
2.80
.53
2.72
.31
Women
2.91
.45
2.80
.54
2.49
.51
3.67
.78
3.55
.68
2.98
.53
Men
3.37
.42
3.51
.56
2.87
.31
Women
3.79
.86
3.56
.73
3.03
.59
4.12
.82
3.63
.81
2.91
1.02
Men
3.33
.90
2.94
.90
2.09
.86
Women
4.36
.66
3.85
.68
3.16
.96
3.20
1.02
3.09
.89
2.71
.70
VE High
MP Low
MP High
Overall
Presence
115
Hypotheses
Estimated pushups. H4 proposed that a VE treatment would be more effective
than an MP treatment in increasing participants‘ estimates of how many pushups they
could perform, and RQ3 inquired about the role of pre-existing self-efficacy. First, a 2 by
2 ANOVA was performed to determine whether or not there were existing differences
between the treatment groups. Although a significant difference was expected between
low and high self-efficacy participants, no differences should exist between the VE and
MP treatments, which would indicate that assignment across groups was not equal.
Results indicated that although there was a difference in Time 2 estimated pushups
between high (M = 24.86, SD = 8.27) and low (M = 7.97, SD = 6.21) self-efficacy
participants, F(1, 48) = 67.48, p < .0005, partial η2 = .58, there was no significant
difference between the MP (M = 14.31, SD = 10.92) and VE (M = 15.55, SD = 11.19)
groups, F(1, 48) = .41, p > .05, partial η2 = .01, nor was there an interaction effect,
F(1, 48) = .10, p > .05, partial η2 = .00.
A mixed design ANOVA was then performed with treatment and pre-existing
self-efficacy as between-subjects variables and participants‘ Time 2 and Time 3
estimated pushups as the within-subjects variables.6 Overall, women demonstrated a
significant increase between Time 2 (M = 13.71, SD = 10.31) and Time 3 (M = 14.36,
SD = 10.28) estimations, F(1, 38) = 7.60, p < .01, partial η2 = .17. There was also an
effect of treatment wherein those in the VE condition experienced a greater increase in
their estimated pushup ability after treatment than those in the MP condition,
F(1, 38) = 7.10, p = .01, partial η2 = .16. There were no differences between low and high
self-efficacy participants over time, F(1, 38) = .06, p > .05, partial η2 = .00, nor was there
116
an interaction effect of treatment and self-efficacy over time, F(1, 38) = 1.27, p > .05,
partial η2 = .03. Figure 16 depicts these results.
Figure 16. Women‘s estimated pushups before and after treatment.
For men, the only significant effect was an increase between estimations at Time
2 (M = 36.24, SD = 11.98) and Time 3 (M = 39.79, SD = 13.55), F(1, 25) = 7.62, p = .01,
partial η2 = .23. There were no effects for treatment, F(1, 23) = .80, p > .05, partial
η2 = .03, self-efficacy, F(1, 23) = 2.44, p > .05, partial η2 = .09, or an interaction,
F(1, 23) = .05, p > .05, partial η2 = .00.
Specific self-efficacy. H5 suggested that the VE stimulus would increase
participants‘ self-reported specific self-efficacy more than the MP treatment, and RQ4
questioned whether pre-existing self-efficacy would interact with the stimulus type. For
women, there were no differences between VE and MP treatments, F(1, 53) = .48,
117
p > .05, partial η2 = .01, nor was the interaction effect significant, F(1, 53) = .19, p > .05,
partial η2 = .00. Only the main effect for pre-existing self-efficacy was significant;
unsurprisingly, high self-efficacy women reported greater feelings of self-efficacy than
low self-efficacy women, F(1, 53) = 48.55, p < .0005, partial η2 = .48. For men, the same
pattern was found; there were no differences between VE and MP treatments,
F(1, 30) = 1.24, p > .05, partial η2 = .04, nor was the interaction effect significant,
F(1, 30) = .09, p > .05, partial η2 = .00. Only the main effect for pre-existing self-efficacy
was significant; high self-efficacy men reported greater feelings of self-efficacy than low
self-efficacy men, F(1, 30) = 20.28, p < .0005, partial η2 = .40.
Actual pushups. H6 predicted that participants receiving the VE treatment would
have greater increase in actual pushup performance over participants who received the
MP treatment; RQ5 questioned the role of pre-existing self-efficacy. A repeatedmeasures ANOVA was performed to assess changes before and after the treatment.7 For
women, the only significant effect was an increase in pushups performed from Time 1
(M = 15.31, SD = 11.10) to Time 3 (M = 19.02, SD = 10.39), F(1, 38) = 12.99, p = .001,
partial η2 = .26. There were no effects for treatment type, F(1, 38) = 1.28, p > .05, partial
η2 = .03, self-efficacy, F(1, 38) = .07, p > .05, partial η2 = .00, or an interaction,
F(1, 38) = 1.81, p > .05, partial η2 = .05. For men, the only significant effect was an
increase in actual pushups performed from Time 1 (M = 40.28, SD = 12.50) to Time 3
(M = 46.14, SD = 13.30), F(1, 25) = 6.08, p < .05, partial η2 = .20. There were no effects
for treatment, F(1, 25) = 1.33, p > .05, partial η2 = .05 or self-efficacy, F(1, 25) = 2.67,
p > .05, partial η2 = .10. The interaction bordered on significance, F(1, 25) = 3.87,
p = .06, partial η2 = .13. The general trend was that participants experienced greater
118
performance in all conditions except the high self-efficacy MP group, which declined.
However, the Levene‘s test was significant for the between-subjects factor, and given the
small and unequal cell sizes, this finding must be considered with caution.
Resemblance. RQ6 inquired about potential differences in perceived resemblance
across conditions. For women, there were main effects for resemblance, but no
interaction effect. Women in the MP condition believed their self-representation
resembled them more than women in the VE condition, F(1, 53) = 54.99, p < .0005,
partial η2 = .51. High self-efficacy women reported higher levels of resemblance than low
self-efficacy women, F(1, 53) = 10.72, p < .005, partial η2 = .17, as can be seen in Figure
17. There was no interaction effect, F(1, 53) = .21, p > .05, partial η2 = .01.
Figure 17. Women‘s perceived resemblance by Condition.
119
For men, only the main effect for stimulus was significant. Men in the MP
condition (M = 3.36, SD = .57) reported their self-representations resembled them more
than men in the VE condition (M = 2.49, SD = .61), F(1, 30) = 5.23, p = .001, partial
η2 = .32, as seen in Figure 18. There were no differences between high self-efficacy
(M = 2.78, SD = .85) and low self-efficacy (M = 2.85, SD = .67) men, F(1, 30) = .00,
p > .05, partial η2 = .00, nor was the interaction effect significant, F(1, 30) = .01, p > .05,
partial η2 = .00.
Figure 18. Men‘s perceived resemblance by Condition.
Identification. RQ7 addressed identification. Again, for women, both main effects
were significant but the interaction was not. Women in the MP condition identified
significantly more with their self-representation than women in the VE condition,
F(1, 53) = 38.84, p < .0005, partial η2 = .42. High self-efficacy women experienced
120
significantly more identification than low self-efficacy women, F(1, 53) = 5.29, p < .05,
partial η2 = .09. The interaction effect was not significant, F(1, 53) = .21, p > .05, partial
η2 = .01. See Figure 19.
Figure 19. Women‘s identification by Condition.
For men, only the main effect for stimulus was significant. Men in the MP
condition reported greater feelings of identification than men in the VE condition,
F(1, 30) = 4.91, p < .05, partial η2 = .14. There were no differences between high selfefficacy and low self-efficacy men, F(1, 30) = .24, p > .05, partial η2 = .01, nor was the
interaction effect significant, F(1, 30) = 2.84, p = .10, partial η2 = .09; see Figure 20.
121
Figure 20. Men‘s identification by Condition.
Presence.5 RQ8 addressed presence. For women, only the main effect for stimulus
was significant. Women in the MP condition (M = 3.08, SD = .74) reported greater
feelings of presence than women in the VE condition (M = 2.29, SD = .55),
F(1, 53) = 8.49, p < .0005, partial η2 = .26. There were no differences between high selfefficacy (M = 2.85, SD = .84) and low self-efficacy (M = 2.71, SD = .72) women,
F(1, 53) = .24, p > .05, partial η2 = .04, nor was the interaction effect significant,
F(1, 53) = .50, p > .05, partial η2 = .01; see Figure 21.
122
Figure 21. Women‘s presence by Condition.
For men, neither of the main effects were significant. Men in the MP condition
(M = 2.63, SD = .62) did not differ from men in the VE condition (M = 2.62, SD = .52) in
feelings of presence, F(1, 30) = .68, p > .05, partial η2 = .02. There were also no
differences between high self-efficacy (M = 2.51, SD = .60) and low self-efficacy
(M = 2.69, SD = .53) men, F(1, 30) = 2.56, p > .05, partial η2 = .08. The interaction
effect was significant, F(1, 30) = 5.57, p < .05, partial η2 = .16. Followup pairwise
comparisons with Bonferroni corrections were conducted; a conservative test was chosen
given the number of statistical tests run. No significant differences were found between
the four groups, however (MP-Low SE, M = 2.87, SD = .31; VR-High SE, M = 2.72,
SD = .31; VR-Low SE, M = 2.57, SD = .62; MP-High SE, M = 2.09, SD = .86).
123
Discussion
This study served as an exploratory examination of the utility of mental imagery
as compared to virtual environments as stimuli for behavior change. Overall, there were
no differences in behavioral outcomes between conditions; however, men and women
both increased the number of pushups achieved after treatment, so it may be that these
treatments were equally effective in enhancing exercise performance. One interesting
finding is that for women, seeing themselves succeed in the virtual environment was
more effective than visualizing their success in boosting the estimated amount of pushups
they believed they could perform. In essence, this represents an effective shift in selfefficacy: women in the VE condition experienced more change in their perceived pushup
efficacy than women in the MP condition. Men, however, did not experience this change.
It may be that because women are generally not as proficient at performing pushups as
men, mere visualization was not able to overcome women‘s perceptions of self-efficacy;
however, seeing themselves perform this behavior in a VE did overcome some of their
reservations. This demonstrates that VEs may be more effective than mental practice for
some groups in promoting health beliefs.
To my knowledge, this is the first study that has attempted to compare the utility
of a computer-generated virtual human and an imagined human in achieving a healthrelated outcome, although many health treatments incorporate VEs or mental imagery.
Several differences were found between conditions on variables that have played a role in
previous studies, such as resemblance, identification, and presence, and these findings
should help guide the design of future experiments comparing virtual and imagined
stimuli. It is unsurprising that there were differences in resemblance between the mental
124
practice and virtual environment conditions; due to limits on the technology, virtual
humans can only resemble the self so closely, and situational variables may influence
differences in appearance (e.g., what clothes a person is wearing or the way hair is
styled). In mental imagery it is easier to generate an identical representation of the self.
Participants also experienced more identification with their representations in the mental
practice condition. It may be that conjuring an image of oneself is a familiar experience,
like daydreaming, in which one projects the self into an imagined world. Thus,
identification may be felt more readily when visualizing. In this virtual environment,
participants had no control over the image; they were merely passive observers, which
also may have hindered their feelings of identification with the virtual self. Future
research should investigate whether these variables play a role in behavioral outcomes,
and if so, what steps can be taken for virtual environments to achieve levels similar to
those of mental imagery.
For presence, it was interesting that only women experienced a difference
between mental imagery and the virtual environment; men experienced equal levels of
immersion and engagement in both conditions. Previous research has noted sex
differences in presence within virtual environments (Nowak et al., 2008). These
differences may be explained by the fact that men play considerably more video games
than women, and women may be less interested and less likely to be engaged in virtual
environments as a result. It is worth noting, however, that VEs were more effective than
mental practice in promoting women‘s self-efficacy, so it is possible that the experience
and effects of presence is different for women. Future research should consider both sex
125
and experience with virtual environments such as video games when considering
presence as a factor.
One issue with this study was that there was no true control without treatment to
compare the effects of VE and MP. Both men and women experienced increases in
pushup performance between Time 1 and Time 3 as well as estimated performance
before and after the stimulus, but without a control condition these results could be
contributed to 1) the tendency to underestimate performance initially so as to not appear
to fail; or 2) a mere presence effect, wherein the presence of the experimenter motivated
participants to exert more effort and exercise harder than they did on their own in the
pretest. Additionally, because participants performed pushups before coming into the lab
for the experiment, a certain sensitization must be considered; participants may have
performed better at Time 3 simply from having done the exercise at Time 1. Thus, no
conclusions can be drawn about the effectiveness of either of these treatments without
replication.
126
CHAPTER TWELVE: CONCLUSION
Summary of Findings
Study 1 confirmed previous studies‘ findings and revealed a mechanism behind
the effectiveness of virtual selves. When observers feel that a virtual self resembles them
physically, they experience more identification with the virtual self. Identification then
bolsters behavioral modeling. Because a virtual self is more likely to resemble the
observer, it is a more potent model than virtual others.
Given that we observed an increase in behavioral modeling based on
identification, it is interesting that there was no effect on self-efficacy in Study 1.
Identification with a model is often hypothesized to enhance feelings of self-efficacy,
which in turn impacts behavior. Study 1 showed that identification and resemblance were
unique predictors of behavior, however. It is possible that the self-efficacy measure was
not specific enough to the behavior to capture any changes; in Study 2, the specific selfefficacy measure did not vary by treatment, although there were changes in women‘s
estimated exercise performance, which can also be interpreted as a measure of selfefficacy (i.e., belief in one‘s ability to perform a behavior at a particular level given the
situational factors).
Study 2 demonstrated the potential of virtual environments for promoting health
behaviors compared to a more traditional method, mental imagery. Given the differences
between pretest and posttest measures, there was some evidence that both types of
treatment were effective in promoting self-efficacy (as measured by estimated behavioral
performance) as well as actual exercise performance. Additionally, VEs were more
effective than mental imagery in promoting women‘s pushup self-efficacy. Study 2 also
127
served an exploratory function, comparing virtual environments to mental imagery on
variables that may be relevant to behavioral outcomes such as resemblance, presence, and
identification. Future studies should continue to explore the strengths and weaknesses of
each of these environments so that maximally effective health promotion stimuli can be
created and implemented in behavior change efforts.
Limitations
As mentioned previously, one limitation of these studies is that the body is not
modeled to resemble the participant, only the face. Although most participants still
identified the body as resembling theirs, future research should incorporate body
modeling as well to create the most accurate form of the virtual self.
Beyond the accuracy of the virtual representations, their realism was also not
considered as a potential factor. The exercise-related pretests used less realistic and
detailed virtual bodies and showed clear effects for virtual selves, whereas the
dissertation studies used more realistic bodies, and the effects were diminished or
nonexistent. It could be that using less realistic bodies directs more attention to the face
and thus participants are more attuned to the self-resemblance, leading to more powerful
effects. Another possibility is that with increasing realism of the stimulus, participants
become more aware of what is not realistic. With a more cartoonish or unreal body, users
may not be as critical of the depiction of exercise or weight gain or loss because their
expectations are not as high. With a highly realistic body, their expectations may be very
high, and if any other aspects of the VE are mismatched with that level of realism, these
discrepancies may stand out and limit the experience of immersion. Determining the role
128
of realism within these treatments is crucial so that designers know where they need or do
not need to divert resources and attention.
Study 1 manipulated the race of the model to match or mismatch the participant.
Although race/ethnicity has been used to examine the comparative effectiveness of
models in previous studies (Anderson & McMillion, 1995; Appiah, 2001; Kalichman et
al., 1995; Pitts et al., 1989) and we did not identify a difference in racism across groups
in Study 1, its potency should not be ignored. Some research has demonstrated that
people react differently to virtual people of a different race (Dotsch & Wigboldus, 2008;
Groom et al., 2009; McCall, Blascovich, Young, & Persky, 2009). Also, Whites may
have been less sensitive to the matched-race condition because they tend not to identify
as part of a specific ethnic group and do not place importance on their racial identity
(Jaret & Reitzes, 1999; Phinney, 1992); unfortunately, there were too few participants in
Study 1 to perform meaningful comparisons across races and conditions. Further analyses
of race should also include a measure of ethnic identification (e.g., Affirmation and
Belonging Scale, Phinney, 1992), which has been shown to be a factor in related studies
(Appiah, 2001).
Study 2 had an issue with participant discrepancy. A pre-experimental pilot
survey conducted with a class participant pool determined that 30 pushups for women
and 60 pushups for men would be appropriate to divide participants into high and low
self-efficacy groups for analysis. Because this was self-report data, one possibility is that
participants merely guessed and overestimated their abilities, which is why our
determined cut-off values proved to be too high for subsequent participants. Another
consideration is that the class that completed the pre-experimental pilot survey had a
129
large student-athlete population (33% of the pretest respondents indicated playing a sport
for Stanford), whereas most of the participants were not run until the following quarter
using mostly a paid pool. Given that Stanford is one of the top performing universities
athletically as well as academically, it is likely that the second group of participants had
considerably lower athletic ability and upper body strength than the pilot group. Despite
specific recruitment efforts to identify high self-efficacy participants (e.g., targeting
campus gyms for advertising, employing snowball recruitment techniques to garner
athletes through various sports teams, advertising specifically for participants with upper
body strength), many of those recruited did not appear for their scheduled experiments
and thus it was not possible to fill the high self-efficacy cells of the experimental design.
(Indeed, only 28% of participants showed up for their experimental timeslots in the last
three weeks of the study.) A replication of this study either changing the pushup quotas or
incorporating a high self-efficacy population may have brought more compelling results.
A final issue is that there were a greater number of women than men who participated in
the study. Between a lack of high self-efficacy participants and a limited number of men,
some cell sizes were small, giving little power for the statistical analyses to find effects.
Another issue with Study 2 is that the mental imagery task asked that participants
imagine themselves from the third person. Perspective plays a role in mental imagery,
and first person imaging (conjuring imagery from one‘s own natural perspective) has
been shown to be more effective than third person imaging (conjuring imagery of the self
from outside of the body) for some tasks (Hardy & Callow, 1999; White & Hardy, 1995).
This fact was considered in the study design, but there was no efficient way to replicate
the first-person perspective in the virtual world so that participants could view the virtual
130
self. A virtual mirror would still provide a third-person perspective of the body, even if
allegedly from a first-person viewpoint. Pushups from the first-person perspective would
have merely shown a virtual floor coming closer to the face and going farther away, a
stimulus that would have been disinteresting and possibly nauseating. Additionally, since
the design called for observation, the sense of looking down and having one‘s virtual
body engage in exercise but experiencing the disconnect of not actually having one‘s
physical body move may have led to diminished presence and caused participants to feel
unsettled and disembodied. Thus, the stimulus was framed from the third person. Future
studies can consider creative ways to overcome this in exercise contexts.
Another issue of using mental practice is that there is no way to control what the
participant actually envisions. Despite the instructions, it is possible that participants did
not follow instructions or that they did not visualize themselves actually succeeding at the
task. Imagery of failure has been shown to have a negative impact on motor tasks
(Powell, 1973; Woolfolk, Parrish, & Murphy, 1985). Thus, if participants had trouble
imagining themselves succeeding at the task, it may have negatively influenced their
physical performance. Negativity may have disproportionately affected low self-efficacy
participants, who already held beliefs that they could not achieve that level of
performance.
Future Directions and Implications
The results of this line of research are compelling, yet they leave many questions
unanswered. Why are virtual selves and virtual humans than resemble the self more
effective models?
131
Although the manipulation in Study 1 was not successful, conceptually the
arguments were supported, and thus require further exploration. To what degree does a
virtual model need to resemble the self to encourage modeling behaviors? A follow up
study might employ a morphing software to blend participants‘ photos at different grades
with others‘ to see if, for example, a model that is 75% similar to the self is more
effective than a model that is 50% or 25% comprised of the self. Another study could
examine what features are crucial to match to promote modeling. For example,
participants may not be influenced whether or not the mouth resembles theirs, but they
may require a similar face shape or skin tone. This research could also be used to further
inform the development of face modeling software for use in creating self-avatars.
The exercise in Study 2, pushups, may have been too difficult to observe changes.
A ceiling effect was observed in that many participants were not able to perform even a
single pushup. Study 2 might be effectively re-run using a different exercise variable with
more variability in performance such as a treadmill test, cycling, or a step test (Bruce,
1974).
Pretest 4 indicated that in some cases, different virtual representations may
actually create different physiological responses. This study examined stationary and
exercising selves; various versions of the virtual self should be created and the bodily
response studied. Physiological data from the studies where people observed the self and
other fattening or slimming would contribute to our understanding of the findings for
Pretests 1 and 2 and Studies 1 and 2. It would also be interesting to study how people
respond to increasingly photorealistic versions of the self, or virtual selves with more or
less of the physical self‘s features (e.g., by blending or morphing the self with another).
132
Another interesting comparison would be how people responded to controlling a virtual
self as an avatar as opposed to watching a virtual self as an agent controlled by an
algorithm. In addition to heart rate and skin conductance, future studies should explore
brain imaging techniques such as fMRI (Baumann et al., 2003; Mraz et al., 2003). New
studies could help determine how people react physiologically to the virtual self, and
whether these reactions are more similar to how they react to computer-generated virtual
humans, photographs of the self, or mirror images.
Longitudinal studies regarding the use of virtual selves as persuasive models are
also necessary. How long can we expect these modeling effects to last? There may be a
point where exposure to a virtual self is no longer compelling. Alternatively, it could be
that if treatment is administered over the long-term, it needs to change over time or be
framed in a more interesting manner (i.e., as a game, or with other participants) to keep
participants engaged so they do not get bored and quit (Madsen, Yen, Wlasiuk, Newman,
& Lustig, 2007). Another consideration is that users may experience reactance at the
thought of a virtual version of themselves being used to control their behaviors.
Discrepancy in the virtual and physical self should also be explored; long-term studies
could also examine what happens if the virtual self experiences positive changes and
rewards while the physical self adheres to an exercise plan, but does not.
Replications of these studies should also be conducted with different populations.
First Lady Michelle Obama recently announced that the focus of her 2010 initiatives
would be childhood obesity. Children and adolescents would be optimal groups for future
studies given the popularity of video games among this age group. Additionally, child
and adult participants of different body types, particularly those deemed overweight or
133
obese, should be recruited. Also, certain groups have been identified as more sedentary or
more at risk for the consequences associated with physical activity, such as heart disease
and diabetes (King, Castro, Wilcox, Eyler, Sallis, & Brownson, 2000; King, Rejeski, &
Buchner, 1998). Certain age or ethnic groups may be targeted for future testing of
exercise interventions using virtual selves.
Virtual selves should also be tested outside of fully immersive virtual
environments as well as laboratory treatments. Desktop and mobile applications should
be developed to test the effectiveness of these treatments in less immersive settings. Also,
these extensions would allow for implementation of the stimuli in more naturalistic
environments, such as the home, workplace, school, or gym. Future research involving
the widespread use of virtual selves as motivators of physical activity should consider
how incorporation could be optimized at micro-, meso- and macro-levels of the
surrounding environment (King, Stoklos, Talen, Brassington, & Killingsworth, 2002).
Identification, self-efficacy, and vicarious reinforcement were examined in this
line of studies, demonstrating that social cognitive theory is a fruitful framework that
could continue to inspire permutations of this research. For example, virtual selves could
easily enable the illustration of both proximal and distal health goals (Bandura & Schunk,
1981; Gollwitzer & Oettingen, 1998). A morbidly obese individual could see herself after
a ten-pound weight loss or after a hundred-pound weight loss. A smoker could compare
his appearance five years and twenty years after quitting smoking or continuing to smoke.
Future research could examine how effective the portrayal of proximal versus distal goals
may be in short-term and long-term treatments. Another use of virtual self-models could
be in the self-regulation of an existing health behavior (Bandura, 2005; Karoly, 1993).
134
These models could be implemented in self-monitoring as reminders to exercise, or as
encouragement to maintain an exercise program or, alternatively, discouragement to
abandon it (Febbraro & Clum, 1998).
The studies conducted here were framed within the perspective of social cognitive
theory, but other perspectives may offer more insight into the mechanisms. Social
comparison theory (Festinger, 1954) may play a role in how we interpret virtual selves in
contrast to virtual others. The theory suggests we seek social cues from others to
determine our own relative success and status and has been used to explain individuals‘
health attitudes and behaviors (Buunk, Zurriaga, Gonzalez, Terol, & Roig, 2006; Gerrard,
Gibbons, Lane, & Stock, 2005; Klein & Cerully, 2007). In these studies, participants may
have been engaging with social comparison with the virtual humans. They may have
employed upward social comparison with thinner avatars or exercising avatars, because
they were exhibiting socially rewarding behaviors, whereas downward social comparison
may have occurred with heavier or inactive avatars. These comparison processes may
have motivated participants‘ reactions, particularly when viewing other avatars as
opposed to the virtual self.
Self-discrepancy theory (Higgins, 1987) is also a useful framework that could be
adopted when examining the effects of virtual selves. The theory suggests that
individuals maintain an actual self, or who you believe yourself to be; the ideal self, or
who you or others want to be; and the ought self, or who you or others think you should
be (Higgins, 1987; Higgins, Roney, Crowe, & Hymes, 1994). This conflict is particularly
relevant to the incorporation of healthful behaviors in a normal lifestyle, for example,
when the actual self is sedentary and out of shape, but the ideal self is healthy and fit.
135
Alternatively, the actual self may be a smoker, but the ought self is a nonsmoker because
of the risk of cancer and the threat secondhand smoke presents to others. Individuals are
motivated by discrepancies between these three versions of the self to make changes in
their health attitudes and behaviors to achieve more synchrony and cohesion (Higgins,
Vookles, & Tykocinski, 1992). Virtual selves may be used to provide a visual
representation of these forms of the self and add incentives for resolving the
discrepancies in a healthful manner.
The concept of wishful identification may be included alongside measures of
identification in the future (Feilitzen & Linne, 1975; Hoffner, 1996; Hoffner &
Buchanan, 2005). People may be differentially influenced by what they can identify with
presently and what they desire to be or what they feel they could be in the future. When
virtual selves are altered, for example, to lose weight, this construct might be useful in
explaining the model‘s effectiveness. Some thin people, for example, may not experience
wishful identification with a slimming model, whereas overweight observers may.
Another explanation is that virtual reality treatments may simply be more
entertaining and fun. Beyond keeping users immediately involved, entertainment value
may promote identification, presence, and long-term adherence (Vorderer, 2000;
Vorderer, Bryant, Pieper, & Weber, 2006; Vorderer, Klimmt, & Ritterfeld, 2004). Future
research could explore the utility of creating virtual exercise gaming environments that
feature photorealistic virtual selves. Reeves and Read (2010) make a case for the
incorporation of gaming and game-like interfaces in the workplace because they promote
engagement and healthy competition. Indeed, similar interfaces could be developed for
individuals to compete with others in health-based games. For example, individual
136
members of a family could log in to a system and record their exercise and diet and be
virtually rewarded for eating the least junk food or getting the most exercise on a given
day.
Were these virtual selves to be incorporated into video games or other
applications, another factor to be explored is the role of narrative. Narratives have been
shown to facilitate transportation, or immersion, into a mediated experience (Green &
Brock, 2000). Indeed, Schneider, Lang, Shin, and Bradley (2004) reported that providing
game players with a story line resulted in greater identification with the character, a
greater sense of presence, and higher physiological arousal. Narratives have also been
shown to be useful in conveying health-related information (Green, 2006; Morgan,
Movius, & Cody, 2009). Dunlop, Wakefield, and Kashima (2008) argue that health
narratives may stimulate self-referent emotions, making them particularly effective in
conveying risk. Thus, these findings indicate that adding a narrative might promote
engagement, arousal, and more thoughts about one‘s own risk; incorporating a virtual self
within the narrative may bolster these experiences. Further investigation should also
determine whether a fun, escapist narrative (e.g., you must exercise and train to prepare
for an upcoming spy mission) or a realistic narrative (e.g., you must exercise and train for
an upcoming marathon) is more engaging and effective for different individuals.
In these studies, virtual selves have been used exclusively as models for exercise
and diet. Future investigations could explore the utility of these models in other health
domains, including smoking cessation, drug use prevention, safe sex practices, or
management of an existing condition such as diabetes. Virtual selves may be
incorporated in existing virtual reality exposure treatments for phobias or social anxiety
137
reduction. Athletes were previously limited to watching videotapes of others‘
performances of a technique or of the self demonstrating a potentially imperfect
technique, but virtual selves could show the self performing a flawless volleyball set, a
perfect roller derby hit, or the ideal baseball swing. These representations may also be
effective in bolstering self-efficacy and perhaps performance in other domains, such as
work tasks or social interactions. For example, people could learn from their virtual
selves how to negotiate, speak assertively, or demonstrate leadership skills. Additionally,
persuasive virtual selves may be used to enhance prosocial behaviors, such as recycling,
conservation, donating blood, or volunteering.
Although these suggestions all describe situations in which virtual selves may be
used to promote positive behaviors, there are of course situations in which these could be
used to manipulate users. The incorporation of virtual selves in persuasive messages calls
for investigation of this concept from an ethical perspective. Previous research has shown
that incorporating elements of the self can lead to subconscious persuasion. If an agent
imperceptibly mimics a person‘s nonverbal gestures, people are more persuaded by the
agent (Bailenson & Yee, 2005). Blending a photograph of the self with that of a political
candidate makes undecided voters prefer that candidate (Bailenson et al., 2006).
Participants who saw themselves associated with a fictitious soft drink brand in an online
advertisement showed preference for that brand later on (Ahn & Bailenson, 2009).
Although the studies reported here sought to have a positive impact on health behaviors,
the same techniques may be used for ill gain. For example, companies could use virtual
selves to sell harmful products, from junk food to cigarettes to diet pills, by creating a
subconscious brand preference through self-association. Given the prevalence of users
138
portraying their photographs on public social networking sites, and many of these sites
(e.g., Facebook) retaining the right to redistribute and use these photographs, it may be
only a matter of time before images of the self become a new currency among
advertisers, political campaigns, and other organizations. Thus, further examination of the
social, psychological, and even legal ramifications of the virtual self is necessary.
Collectively, these studies have provided a wealth of information on the utility of
virtual selves as models for health behavior change. Within the framework of social
cognitive theory, virtual selves can be used to promote identification, demonstrate
vicarious reinforcement, and enhance feelings of self-efficacy regarding health behaviors.
Given the amount of time people spend engaging with new media via computers, gaming
systems, and mobile devices, these models have the potential to become powerful
motivators in day-to-day health efforts. As we observe our virtual selves evolve as they
adopt and maintain healthful behaviors, we, too, can become healthy individuals.
139
NOTES
1
Some of the pretests outlined here have been previously published in journals.
Pretests 1, 2, and 3 appeared in Media Psychology (Fox & Bailenson, 2009). Pretest 5
appeared in PRESENCE: Teleoperators & Virtual Environments (Fox et al., 2009).
2
An initial sample of N = 97 was obtained; due to a technological failure, some
participants‘ data did not record in two conditions. A random sample was taken from the
unaffected condition to balance the number of participants in each condition, leaving a
final sample of N = 63. The statistical models perform similarly with or without the extra
participants.
3
In this experiment, perspective was also manipulated in that participants saw the
treatment from either a first or third person perspective. No differences were found
between the two on any of the dependent variables, and thus perspective is not discussed
further.
4
This finding is also significant using the data without the wisterization.
5
All tests involving the presence variable were repeated using the subscales of
environmental, social, and self-presence. However, no distinct findings were found using
the subscales, and so analyses are reported for the summary variable.
6
A percent change variable was also calculated by subtracting the participants‘
Time 2 estimated pushups from the Time 3 estimated pushups and dividing by the Time 2
estimated pushups. Using this measure, women in the VE condition (M = .29, SD = .55)
reported a greater increase in the number of pushups they believed they could do than
women in the MP condition (M = .00, SD = .22), F(1, 38) = 6.30, p < .05, partial η2 = .14.
There were no differences between high self-efficacy (M = .05, SD = .13) and low self-
140
efficacy (M = .10, SD = .46) women, F(1, 38) = 2.42, p > .05, partial η2 = .06, nor was
the interaction effect significant, F(1, 38) = 2.40, p > .05, partial η2 = .06. There were no
significant effects for men. There were no differences between men who received the VE
(M = .15, SD = .31) or MP (M = .08, SD = .28) treatment on estimated pushups,
F(1, 25) = .09, p > .05, partial η2 = .00. There were no differences between high selfefficacy (M = .23, SD = .36) and low self-efficacy (M = .05, SD = .24) men,
F(1, 25) = 2.38, p > .05, partial η2 = .09, nor was the interaction effect significant,
F(1, 25) = .47, p > .05, partial η2 = .02.
7
A percent change was also calculated by subtracting the participants‘ Time 3
pushups from the Time 1 pushups and dividing by the Time 1 pushups. There were no
significant effects for women. There were no differences between women who received
the VE (M = .35, SD = .79) or MP (M = .61, SD = 1.34) treatment percent change of
pushups, F(1, 38) = .06, p > .05, partial η2 = .00. There were no differences between high
self-efficacy (M = .20, SD = .33) and low self-efficacy (M = .71, SD = 1.44) women,
F(1, 38) = .86, p > .05, partial η2 = .02, nor was the interaction effect significant,
F(1, 25) = .67, p > .05, partial η2 = .02. No significant effects were found for men, either.
There were no differences between men who received the VE (M = .18, SD = .28) or MP
(M = .18, SD = .31) treatment percent change of pushups, F(1, 25) = .37, p > .05, partial
η2 = .02. There were no differences between high self-efficacy (M = .07, SD = .30) and
low self-efficacy (M = .23, SD = .27) men, F(1, 38) = 3.07, p = .09, partial η2 = .11, nor
was the interaction effect significant, F(1, 25) = 3.14, p = .09, partial η2 = .11.
141
APPENDIX A: STUDY 1 SURVEY ITEMS
Avatar Resemblance
(1 = Definitely did not look like me at all; 2 = Mostly did not look like me; 3 = Looked
somewhat like me; 4 = Mostly looked like me; 5 = Definitely looked a lot like me)
1. To what degree did the avatar you saw resemble you in the face?
2. To what degree did the avatar you saw resemble you in the body (before any changes)?
3. Overall, how much do you feel the avatar resembled you?
Avatar Race
(African/African-American/Black; Asian/Asian-American; Caucasian/EuropeanAmerican/White; Latino/a; Other)
4. What race did the avatar appear to be?
Identification
(1 = Strongly disagree; 2 = Somewhat disagree; 3 = Neither disagree nor agree; 4 =
Somewhat agree; 5 = Strongly agree)
5. While observing the virtual scene, I felt as if I was part of the action.
6. While observing the virtual scene, I forgot myself and was completely absorbed.
7. I think I have a good understanding of my avatar.
8. While observing the virtual scene, I could feel the experiences my avatar portrayed.
9. While observing the virtual scene, I wanted my avatar to succeed in achieving his or
her goals by exercising and losing weight.
10. While observing the virtual scene, I felt I could really get inside my avatar.
11. At key moments, I felt I knew exactly what my avatar was going through.
12. While observing the virtual scene, I did not want my avatar to fail and gain weight.
142
Presence
(1 = Not at all; 2 = Slightly; 3 = Moderately; 4 = Very much; 5 = Extremely)
13. To what extent do you feel the avatar you saw is an extension of yourself?
14. To what extent do you feel that if something happens to the avatar you saw, it feels
like it is happening to you?
15. To what extent do you feel you embodied the avatar you saw?
16. To what degree did you identify with the avatar you saw?
17. To what extent did you feel you were in the same room as the avatar you saw?
18. To what extent did the avatar you saw seem real?
19. To what extent were you involved in the environment?
20. To what extent did you feel like you were inside the environment?
21. To what extent did you feel surrounded by the environment?
22. To what extent did it feel like you visited another place?
23. How much did the environment seem like the real world?
Specific Self-Efficacy
(1 = Strongly disagree; 2 = Somewhat disagree; 3 = Neither disagree nor agree; 4 =
Somewhat agree; 5 = Strongly agree)
24. I could perform the exercise I saw in the real world at a steady rate for 30 minutes.
25. I would have no problems performing the exercise I saw at a steady rate for 30
minutes.
26. Given the opportunity, nothing would stop me from performing the exercise I saw at
a steady rate for 30 minutes.
27. I would be successful performing the exercise I saw at a steady rate for 30 minutes.
143
28. There are too many obstacles for me to perform the exercise I saw at a steady rate for
30 minutes.
29. I would fail at performing the exercise I saw at a steady rate for 30 minutes.
30. The exercise I saw is too challenging to perform at a steady rate for 30 minutes.
General Self-Efficacy
Please respond to the following items. Consider how you are feeling right now, at this
moment, and whether or not you could do these things continuously for the next three
months. I feel I could do the following activities for the next three months: (1 = Very
confident I could not do it; 2 = Mostly confident that I could not do it; 3 = Not sure if I
could or could not do it; 4 = Mostly confident that I could do it; 5 = Very confident that I
could do it).
31. Stick to my exercise program even after a long, tiring day at work or school.
32. Set aside time for physical activity such as walking, jogging, swimming, biking, or
other continuous forms of exercise for at least 30 minutes 3 times a week.
33. Stick to my exercise program even when I have excessive demands at work or school.
34. Spend less time reading, studying, or working on the computer in order to exercise
more.
35. Get up earlier, give up naps, stay up later, or make other alterations to my sleep
schedule to make time to exercise.
36. Make time, even on weekends, to exercise.
37. Stick to my exercise program even when friends and family are demanding more time
from me.
144
38. Stick to my exercise program even when I have a lot of household chores and errands
to attend to.
39. Exercise even when I am feeling sad or depressed.
40. Stick to my exercise program even when social obligations are very time consuming.
41. Stick to my exercise program when undergoing a stressful event (e.g., finals, moving,
a death in the family).
42. Stick to an exercise program even when the results I want are taking a long time to
achieve.
43. Stick to an exercise program even when I‘m not seeing the results I want.
44. Exercise even though it is physically challenging.
45. Stick with an exercise program even if I don‘t reach a specific goal I have.
Racism
(1 = Strongly disagree; 2 = Somewhat disagree; 3 = Neither disagree nor agree;
4 = Somewhat agree; 5 = Strongly agree)
46. It is easy to understand the anger of minorities in America.
47. The streets are not safe these days without police or other surveillance.
48. Minorities are getting too demanding in their push for equal rights.
49. Over the past few years, minorities have gotten more economically than they deserve.
50. Over the past few years, the government and media have shown more respect to
minorities than they deserve.
51. Minorities are gaining more advantage than they deserve through affirmative action
in schools and workplaces.
145
APPENDIX B: STUDY 1 CONDITION x SEX FACTORIAL ANOVA RESULTS
DV
Identification
Specific SelfEfficacy
Effect
F-test Results
Conclusion
Condition
F(2, 69) = .55, p > .05, partial η2 = .02
No effect
Sex
F(2, 69) = .02, p > .05, partial η2 = .00
No effect
Interaction
F(2, 69) = 2.31, p = .11, partial η2 = .06
No effect
Condition
F(2, 69) = .70, p > .05, partial η2 = .02
No effect
Sex
F(2, 69) = .59, p > .05, partial η2 = .01
No effect
Interaction
F(2, 69) = 2.71, p = .07, partial η2 = .07
Borderline
effect
Exercise
Repetitions
Presence
Condition
F(2, 69) = .83, p > .05, partial η2 = .02
No effect
Sex
F(2, 69) = 1.36, p > .05, partial η2 = .02
No effect
Interaction
F(2, 69) = .65, p > .05, partial η2 = .02
No effect
Condition
F(2, 69) = .03, p > .05, partial η2 = .00
No effect
Sex
F(2, 69) = 2.88, p = .09, partial η2 = .04
No effect
Interaction
F(2, 69) = 1.34, p > .05, partial η2 = .04
No effect
Condition
F(2, 69) = 1.14, p > .05, partial η2 = .03
No effect
Sex
F(2, 69) = 8.29, p = .005, partial η2 = .11 Men >
General SelfEfficacy
Women
Interaction
F(2, 69) = .26, p > .05, partial η2 = .01
No effect
146
APPENDIX C: STUDY 1 SELF VS. OTHER (SIMILAR & DISSIMILAR) t-TEST
RESULTS
Self
Other
t-Test
DV
M
SD
M
Conclusion
SD
t(72) = 3.96,
Resemblance
2.76
.66
2.09
.70
p < .0005,
Self > Other
Cohen‘s d = .99
t(72) = .27, p > .05,
Identification
2.89
.83
2.84
.71
No effect
Cohen‘s d = .07
t(72) = .09, p > .05,
Specific SE
3.76
1.08
3.79
.96
No effect
Cohen‘s d = .03
Exercise
t(72) = .17, p > .05,
41.54 40.97
39.83
41.55
No effect
Cohen‘s d = .04
Repetitions
t(72) = .29, p > .05,
Presence
2.87
.72
2.92
.64
No effect
Cohen‘s d = .07
t(72) = 1.44, p > .05,
General SE
3.81
.97
3.49
.88
No effect
Cohen‘s d = .35
147
APPENDIX D: STUDY 1 SELF & SIMILAR VS. DISSIMILAR t-TEST RESULTS
Self/Similar
Dissimilar
t-Test
DV
M
SD
M
Conclusion
SD
t(73) = 6.06, p <
Self/Similar >
Resemblance
2.65
.65
1.72
.58
.0005, Cohen‘s d =
Dissimilar
1.51
t(73) = .62, p > .05,
Identification
2.80
.76
2.92
.74
No effect
Cohen‘s d = .16
t(73) = .95, p > .05,
Specific SE
3.86
1.00
3.63
.98
No effect
Cohen‘s d = .23
Exercise
t(73) = 1.22, p > .05,
44.44 41.89
32.28
38.03
No effect
Cohen‘s d = .30
Repetitions
t(73) = .13, p > .05,
Presence
2.89
.67
2.91
.67
No effect
Cohen‘s d = .02
t(73) = .91, p > .05,
General SE
3.67
1.00
3.46
.70
No effect
Cohen‘s d = .24
148
APPENDIX E: STUDY 2 SURVEY ITEMS
Pretest
Time 1 Pushup Performance
1. Please indicate how many pushups you think you could do in one minute while
maintaining good form. These are full form pushups from your toes, not your knees,
where your back is flat and your butt is not arched in the air. In good form, your body
goes all the way down until nearly touching the ground and is pushed all the way up until
the arms are nearly straight. If you need to take a break, that‘s fine, but do not count any
repetitions after sixty seconds.
How many pushups did you complete in one minute?
Time 1 Pushup Self-Efficacy
How confident are you that you could perform the following number of pushups in one
minute while maintaining good form? (1 = Definitely could not do it; 2 = Probably could
not do it; 3 = Not sure; 4 = Probably could do it; 5 = Definitely could do it)
2. 5 pushups in 1 minute
3. 10 pushups in 1 minute
4. 15 pushups in 1 minute
5. 20 pushups in 1 minute
6. 25 pushups in 1 minute
7. 30 pushups in 1 minute
8. 40 pushups in 1 minute
9. 50 pushups in 1 minute
10. 60 pushups in 1 minute
149
11. 70 pushups in 1 minute
12. 80 pushups in 1 minute
13. More than 80 pushups in 1 minute
Experimental Survey
These questions are about the images and human representation you saw in the first task.
Resemblance
(1 = Definitely did not look like me at all; 2 = Mostly did not look like me; 3 = Looked
somewhat like me; 4 = Mostly looked like me; 5 = Definitely looked a lot like me)
1. To what degree do you feel the representation resembled you in the face?
2. To what degree do you feel the representation resembled you in the body?
3. Overall, how much do you feel the representation resembled you?
Identification
(1 = Strongly disagree; 2 = Somewhat disagree; 3 = Neither disagree nor agree; 4 =
Somewhat agree; 5 = Strongly agree)
4. While observing the scene, I felt as if I was part of the action.
5. While observing the scene, I forgot myself and was completely absorbed.
6. I think I have a good understanding of my representation.
7. While observing the scene, I could feel the experiences my representation portrayed.
8. While observing the scene, I wanted my representation to succeed in achieving his or
her goals by finishing the pushups in one minute.
9. While observing the scene, I felt I could really get inside my representation.
10. At key moments, I felt I knew exactly what my representation was going through.
150
11. While observing the scene, I did not want my representation to fail and not finish the
set of pushups.
Presence
(1 = Not at all; 2 = Slightly; 3 = Moderately; 4 = Very much; 5 = Extremely)
12. To what extent to do you feel the representation is an extension of yourself?
13. To what extent do you feel that if something happens to the representation, it feels
like it is happening to you?
14. To what extent do you feel you embodied the representation you saw?
15. To what degree did you identify with the representation?
16. To what extent did you feel you were in the same room with the representation?
17. To what extent did the representation seem real?
18. To what extent were you involved in the world you saw?
19. To what extent did you feel like you were inside the world you saw?
20. To what extent did you feel surrounded by the world you saw?
21. To what extent did it feel like you visited another place?
22. How much did the world you saw seem like the real world?
Specific Self-Efficacy
(1 = Strongly disagree; 2 = Somewhat disagree; 3 = Neither disagree nor agree; 4 =
Somewhat agree; 5 = Strongly agree)
23. I could perform the exercise I saw in the real world at the pace I saw (that number of
pushups in 1 minute).
24. I would have no problems performing pushups at the pace I saw.
151
25. Given the opportunity, nothing would stop me from performing pushups at the pace I
saw.
26. I would be successful performing pushups at the pace I saw.
27. There are too many obstacles for me to perform pushups at the pace I saw.
28. I would fail at performing pushups at the pace I saw.
29. Pushups are too challenging to perform at the pace I saw.
152
REFERENCES
Aadahl, M., & Jørgensen, T. (2003). Validation of a new self-report instrument for
measuring physical activity. Medicine & Science in Sports & Exercise, 35, 11961202.
Ahn, S. J. & Bailenson, J. N. (2009). My picture is better than yours: Self-endorsing
in online advertisements. Proceedings of the 59th Annual
International Communication Association Conference, May 21-25, Chicago, IL.
Anderson, C. A., & Bushman, B. J. (2001). Effects of violent video games on aggressive
behavior, aggressive cognition, aggressive affect, physiological arousal, and
prosocial behavior: A meta-analytic review of the scientific literature.
Psychological Science, 12, 353-359.
Anderson, C. A., & Dill, K. E. (2000). Video games and aggressive thoughts, feelings,
and behavior in the laboratory and life. Journal of Personality & Social
Psychology, 78, 772-790.
Anderson, C. A., Gentile, D. A., & Buckley, K. E. (2007). Violent video game effects on
children and adolescents: Theory, research, and public policy. New York: Oxford
University Press.
Anderson, E. S., Winett, R. A., & Wojcik, J. R. (2007). Self-regulation, self-efficacy,
outcome expectations, and social support: Social cognitive theory and nutrition
behavior. Annals of Behavioral Medicine, 34, 304-312.
153
Anderson, E. S., Wojcik, J. R., Winett, R. A., & Williams, D. M. (2006). Social-cognitive
determinants of physical activity: The influence of social support, self-efficacy,
outcome expectations, and self-regulation among participants in a church-based
health promotion study. Health Psychology, 25, 510-520.
Anderson, R. B., & McMillion, P. Y. (1995). Effects of similar and diversified modeling
On African-Amercian women‘s efficacy expectations and intentions to perform
breast self-examination. Health Communication, 7, 327-343.
Andsager, J. L., Austin, E. W., & Pinkleton, B. E. (2001). Questioning the value of
realism: Young adults‘ processing of messages in alcohol-related public service
announcements and advertising. Journal of Communication, 51, 121-142.
Andsager, J. L., Bemker, V., Choi, H.-L., & Torwel, V. (2006). Perceived similarity of
exemplar traits and behavior: Effects on message evaluation. Communication
Research, 33, 3-18.
Annesi, J. J., & Mazas, J. (1997) Effects of virtual reality-enhanced exercise equipment
on adherence and exercise-induced feeling states. Perceptual & Motor Skills, 85,
835-844.
Appiah. O. (2001). Ethnic identification on adolescents‘ evaluations of advertisements.
Journal of Advertising Research, 41(5), 7-22.
Armitage, C. J., & Conner, M. (2001). Social cognitive determinants of blood donation.
Journal of Applied Social Psychology, 31, 1431-1457.
154
Atienza, A. A., King, A. C., Oliveira, B. M., Ahn, D. K., & Gardner, C. D. (2008). Using
hand-held computer technologies to improve dietary intake. American Journal of
Preventive Medicine, 34, 514-518.
Babin, L. A., & Burns, A. C. (1998). A modified scale for the measurement of
communication-evoked mental imagery. Psychology & Marketing, 15, 261-278.
Baecke, J. A. H., Burema, J., & Fritjers, J. E. R. (1982). A short questionnaire for the
measurement of habitual physical activity in epidemiological studies. American
Journal of Clinical Nutrition, 36, 936-942.
Bailenson, J. N. & Beall, A. C. (2006). Transformed social interaction: Exploring the
digital plasticity of avatars. In R. Schroeder & A.-S. Axelsson (Eds.), Avatars at
work and play: Collaboration and interaction in shared virtual environments (pp.
1-16). New York: Springer-Verlag.
Bailenson, J. N., Beall, A. C., Blascovich, J., & Rex, C. (2004). Examining virtual busts:
Are photogrammetrically generated head models effective for person
identification? PRESENCE: Teleoperators & Virtual Environments, 13, 416-427.
Bailenson, J. N., Beall, A. C., Loomis, J., Blascovich, J., & Turk, M. (2004).
Transformed social interaction: Decoupling representation from behavior and
form in collaborative virtual environments. PRESENCE: Teleoperators & Virtual
Environments, 13, 428-441.
Bailenson, J. N., Beall, A. C., Loomis, J., Blascovich, J., & Turk, M. (2005).
Transformed social interaction, augmented gaze, and social influence in
immersive virtual environments. Human Communication Research, 31, 511-537.
155
Bailenson, J. N., & Blascovich, J. (2004). Avatars. In W. S. Bainbridge (Ed.), Berkshire
Encyclopedia of Human-Computer Interaction (pp. 64–68). Great Barrington,
MA: Berkshire Publishing Group.
Bailenson, J., Blascovich, J., Beall, A., & Loomis, J. (2003). Interpersonal distance in
immersive virtual environments. Personality & Social Psychology Bulletin, 29, 115.
Bailenson, J. N., Blascovich, J., & Guadagno, R. E. (2008). Self representations in
immersive virtual environments. Journal of Applied Social Psychology, 38, 26732690.
Bailenson, J. N., Garland, P., Iyengar, S., & Yee, N. (2006). Transformed facial similarity
as a political cue: A preliminary investigation. Political Psychology, 27, 373-386.
Bailenson, J. N., Patel, K., Nielsen, A., Bajscy, R., Jung, S.-H., & Kurillo, G. (2008). The
Effect of interactivity on learning physical actions in virtual reality. Media
Psychology, 11, 354-376.
Bailenson, J. N., & Yee, N. (2005). Digital chameleons: Automatic assimilation of
nonverbal gestures in immersive virtual environments. Psychological Science, 16,
814-819.
Bailenson, J. N., & Yee, N. (2007). Virtual interpersonal touch: Haptic interaction and
copresence in collaborative virtual environments. International Journal of
Multimedia Tools & Applications, 37, 5-14.
156
Bailenson, J. N., Yee, N., Blascovich, J., Beall, A. C., Lundblad, N., & Jin, M. (2008).
The use of immersive virtual reality in the learning sciences: Digital
transformations of teachers, students, and social context. Journal of the Learning
Sciences, 17, 102-141.
Baker, C. W., Little, T. D., & Brownell, K. D. (2003). Predicting adolescent eating and
activity behaviors: The role of social norms and personal agency. Health
Psychology, 22, 189-198.
Bandura, A. (1965). Influence of models‘ reinforcement contingencies on the acquisition
of imitative responses. Journal of Personality & Social Psychology, 1, 589-595.
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1982). Self-efficacy mechanism in human agency. American Psychologist,
37, 122-147.
Bandura, A. (1986). Social foundations of thought and action. Englewood Cliffs, NJ:
Prentice-Hall.
Bandura, A. (1989). Human agency in social cognitive theory. American Psychologist,
44, 1175-1184.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Bandura, A. (2000). Health promotion from the perspective of social cognitive theory. In
P. Norman, C. Abraham, & M. Conner (Eds.), Understanding & changing health
behavior (pp. 299-339). Reading, UK: Harwood.
Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology,
3, 265-299.
157
Bandura, A. (2002). Growing primacy of human agency in adaptation and change in the
electronic era. European Psychologist, 7, 2-16.
Bandura, A. (2004). Health promotion by social cognitive means. Health Education &
Behavior, 31, 143-164.
Bandura, A. (2005). The primacy of self-regulation in health promotion. Applied
Psychology: An International Review, 54, 245-254.
Bandura, A. (2006). Toward a psychology of human agency. Perspectives on
Psychological Science, 1, 164-180.
Bandura, A. (2007). Much ado over a faulty conception of perceived self-efficacy
grounded in faulty experimentation. Journal of Social & Clinical Psychology, 26,
641-658.
Bandura, A., & Cervone, D. (1983). Self-evaluative and self-efficacy mechanisms
governing the motivational effects of goal systems. Journal of Personality &
Social Psychology, 45, 1017-1028.
Bandura, A., & Huston, A. C. (1961). Identification as a process of incidental learning.
Journal of Abnormal & Social Psychology, 63, 311-318.
Bandura, A. & Rosenthal, T. L. (1966). Vicarious classical conditioning as a function of
arousal level. Journal of Personality & Social Psychology, 3, 54-62.
Bandura, A., Ross, D., & Ross, S. A. (1963). Vicarious reinforcement and imitative
learning. Journal of Abnormal & Social Psychology, 67, 601-607.
Bandura, A., & Schunk, D. H. (1981). Cultivating competence, self-efficacy, and intrinsic
interest through proximal self-motivation. Journal of Personality & Social
Psychology, 41, 586-598.
158
Baranowski, T., Baranowski, J., Cullen, K. W., Marsh, T., Islam, N., Zakeri, I., et al.
(2003). Squire‘s Quest! Dietary outcome evaluation of a multimedia game.
American Journal of Preventive Medicine, 24, 52-61.
Baranowski, T., Buday, R., Thompson, D. I., & Baranowski, J. (2008). Playing for real:
Video games and stories for health behavior change. American Journal of
Preventive Medicine, 34, 74-82.
Basil, M. D. (1996). Identification as a mediator of celebrity effects. Journal of
Broadcasting & Electronic Media, 40, 478-497.
Baumann, S., Neff, C., Fetzick, S., Stangl, G., Basler, L., Vereneck, R., et al. (2003). A
virtual reality system for neurobehavior and functional MRI studies.
CyberPsychology & Behavior, 6, 259-266.
Baumann, S. B., & Sayette, M. A. (2006). Smoking cues in a virtual world provoke
craving in cigarette smokers. Psychology of Addictive Behaviors, 20, 484-489.
Beaudoin, C. E., Fernandez, C., Wall, J. L., & Farley, T. A. (2007). Promoting healthy
Eating and physical activity: Short-term effects of a mass media campaign.
American Journal of Preventive Medicine, 32, 217-223.
Bem, D. J. (1967). Self-perception: An alternative interpretation of cognitive dissonance
phenomena. Psychological Review, 74, 183-200.
Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in
experimental social psychology (Vol. 6, pp. 1-99). New York: Academic Press.
Bessière, K., Seay, A. F., & Kiesler, S. (2007). The ideal elf: Identity exploration in
World of Warcraft. CyberPsychology & Behavior, 10, 530-535.
159
Bickmore, T., Gruber, A., & Picard, R. (2005). Establishing the computer-patient
working alliance in automated health behavior change interventions. Patient
Education & Counseling, 59, 21-30.
Biocca, F., Harms, C., & Burgoon, J. K. (2003). Toward a more robust theory and
measure of social presence: Review and suggested criteria. PRESENCE:
Teleoperators & Virtual Environments, 12, 456-480.
Blascovich, J. (2001). Immersive virtual environments and social behavior. Science
Briefs: Psychological Science Agenda, 14, 8-9.
Blascovich, J. (2002). Social influence within immersive virtual environments. In R.
Schroeder (Ed.), The social life of avatars: Presence and interaction in shared
virtual environments (pp. 127-145). London: Springer-Verlag.
Blascovich, J., Loomis, J., Beall, A. C., Swinth, K. R., Hoyt, C. L., & Bailenson, J. N.
(2002). Immersive virtual environment technology as methodological tool for
social psychology. Psychological Inquiry, 13, 103-124.
Bock, B. C., Marcus, B. H., Pinto, B. M., & Forsyth, L. H. (2001). Maintenance of
physical activity following an individualized motivationally tailored intervention.
Annals of Behavioral Medicine, 23(2), 79-87.
Botella, C., García-Palacios, A., Villa, H., Baños, R. M., Quero, S., Alcañiz, M., et al.
(2007).Virtual reality exposure in the treatment of panic disorder and
agoraphobia: A controlled study. Clinical Psychology & Psychotherapy, 14, 164175.
Brown, W. J., & Cody, M. J. (1991). Effects of a prosocial televison soap opera in
promoting women‘s status. Human Communication Research, 18, 114-142.
160
Bruce, R. A. (1974). Methods of exercise testing: Step test, bicycle, treadmill, isometrics.
Journal of American Cardiology, 33, 715-720.
Bryanton, C., Bossé, J., Brien, M., McLean, J., McCormick, A., & Sveistrup, H. (2006).
Feasibility, motivation, and selective motor control: Virtual reality compared to
conventional home exercise in children with cerebral palsy. CyberPsychology &
Behavior, 9, 123-128.
Buckworth, J., Lee, R. E., Regan, G., Schneider, L., & DiClemente, C. C. (2007).
Decomposing intrinsic and extrinsic motivation for exercise: Application to stages
of motivational readiness. Psychology of Sport & Exercise, 8, 441-461.
Bucy, E. P., & Tao, C.-C. (2007). The mediated moderation model of interactivity. Media
Psychology, 9, 647-672.
Bussey, K., & Perry, D. G. (1982). Same-sex imitation: The avoidance of cross-sex
models or the acceptance of same-sex models? Sex Roles, 8, 773-784.
Buunk, A. P., Zurriaga, R., Gonzalez, P., Terol, C., & Roig, S. L. (2006). Targets and
dimensions of social comparison among people with spinal cord injury and other
health problems. British Journal of Health Psychology, 11, 677-693.
Cable, D. M., & Yu, K. Y. T. (2006). Managing job seekers‘ organizational image
beliefs: The role of media richness and media credibility. Journal of Applied
Psychology, 91, 828-840.
Campbell, M. K., DeVellis, B. M., Strecher, V. J., Ammerman, A. S., DeVellis, R. F., &
Sandler, R. S. (1994). Improving dietary behavior: The effectiveness of tailored
messages in primary care settings. American Journal of Public Health, 84, 783787.
161
Cassell, J. (2000). Embodied conversational interface agents. Communications of the
ACM, 43(4), 70-78.
Castronova, E. (2004). The price of bodies: A hedonic pricing model of avatar attributes
in a synthetic world. KYKLOS, 57, 173-196.
Castronova, E. (2005). Synthetic worlds: The business and culture of online games.
Chicago: University of Chicago Press.
Chandler, P., & Sweller, J. (1991). Cognitive load theory and the format of instruction.
Cognition & Instruction, 8, 293-332.
Cho, S., Ku, J., Park, J., Han, K., Lee, H., Choi, Y. K., et al. (2008). Development and
verification of an alcohol-craving induction tool using virtual reality: Craving
characteristics in a social pressure situation. CyberPsychology & Behavior, 11,
302-309.
Chuang, T.-Y., Chen, C.-H., Chang, H.-A., Lee, H.-C., Chou, C.-L., & Doong, J.-L.
(2003). Virtual reality serves as a support technology in cardiopulmonary exercise
testing. PRESENCE: Teleoperators & Virtual Environments, 12, 326-331.
Clark, N. M., & Dodge, J. A. (1999). Exploring self-efficacy as a predictor of disease
management. Health Education & Behavior, 26, 72-89.
Coelho, C. M., Santos, J. A., Silvério, J., & Silva, C. F. (2006). Virtual reality and
acrophobia: One-year follow-up and case study. CyberPsychology & Behavior, 9,
336-341.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.).
Hillsdale, NJ: Erlbaum.
162
Cohen, J. (2001). Defining identification: A theoretical look at the identification of
audiences with media characters. Mass Communication & Society, 4, 245-264.
Cohen, J. (2002). Deconstructing Ally: Explaining viewers‘ interpretations of popular
television. Media Psychology, 4, 253-277.
Cote, S., & Bouchard, S. (2005). Documenting the efficacy of virtual reality exposure
With psychophysiological and information processing measures. Applied
Psychophysiology & Biofeedback, 30, 217-232.
Cumming, J. (2008). Investigating the relationship between exercise imagery, leisuretime exercise behavior, and self-efficacy. Journal of Applied Sport Psychology,
20, 184-198.
Daft, R. L., Lengel, R. H. & Trevino, L. K. (1987). Message equivocality, media
selection, and manager performance: Implications for information systems. MIS
Quarterly, 11, 355-366.
Dahlquist, L. M., McKenna, K. D., Jones, K. K., Dillinger, L., Weiss, K. E., &
Ackerman, C. S. (2007). Active and passive distraction using a head-mounted
display helmet: Effects on cold pressor pain in children. Health Psychology, 26,
794-801.
DeBusk, R. F., Miller, N. H., Superko, H. R., Dennis, C. A., Thomas, R. J., Lew, H. T., et
al. (1994). A case-management system for coronary risk factor modification after
acute myocardial infarction. Annals of Internal Medicine, 120, 721-729.
De Leo, G., Ponder, M., Molet, T., Fato, M., Thalmann, D., Magnenat-Thalmann, N., et
al. (2003). A virtual reality system for the training of volunteers involved in
health emergency situations. CyberPsychology & Behavior, 6, 267-274.
163
Deutsch, J. E., & Mirelman, A. (2007). Virtual reality-based approaches to enable
walking for people poststroke. Topics in Stroke Rehabilitation, 14, 45-53.
Dion, K., Berscheid, E., & Walster, E. (1972). What is beautiful is good. Journal of
Personality & Social Psychology, 34, 366-375.
Dotsch, R., & Wigboldus, D. H. J. (2008). Virtual prejudice. Journal of Experimental
Social Psychology, 44, 1194-1198.
Dowrick, P. W. (1999). A review of self modeling and related interventions. Applied &
Preventive Psychology, 8, 23-39.
Driskell, J. E., Copper, C., & Moran, A. (1994). Does mental practice enhance
performance? Journal of Applied Psychology, 79, 481-492.
Dunlop, S., Wakefield, M., & Kashima, Y. (2008). Can you feel it? Negative emotion,
risk, and narrative in health communication. Media Psychology, 11, 52-75.
Eastin, M. S., Appiah, O., & Cicchirllo, V. (2009). Identification and the influence of
cultural stereotyping on postvideogame hostility. Human Communication
Research, 35, 337-356.
Farquhar, J. W., Fortmann, S. P., Flora, J. A., Taylor, B., Haskell, W. L., Williams, P. T.,
et al. (1990). Effects of community-wide education on cardiovascular disease risk
factors: The Stanford Five-City Project. Journal of the American Medical
Association, 264, 359-365.
Farquhar, J. W., Fortmann, S. P., Maccoby, N., Haskell, W. L., Williams, P. T., Flora, J.
A., et al. (1985). The Stanford Five-City Project: Design and methods. American
Journal of Epidemiology, 122, 323-334.
164
Febbraro, G. A. R., & Clum, G. A. (1998). Meta-analytic investigation of the
effectiveness of self-regulatory components in the treatment of adult problem
behaviors. Clinical Psychology Review, 18, 143-161.
Feilitzen, C., & Linne, O. (1975). Identifying with television characters. Journal of
Communication, 25(4), 51-55.
Feltz, D. L. (1984). Self-efficacy as a cognitive mediator of athletic performance. In W.
F. Straub & J. M. Williams (Eds.), Cognitive sport psychology (pp. 191-198).
Lansing, NY: Sport Science Associates.
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2),
117-140.
Fishbein, M., & Cappella, J. N. (2006). The role of theory in developing effective health
communications. Journal of Communication, 56, S1-S17.
Flora, J. A., & Thoresen, C. E. (1988). Reducing the risk of AIDS in adolescents.
American Psychologist, 43, 965-970.
Forsyth, A. D., & Carey, M. P. (1998). Measuring self-efficacy in the context of HIV risk
reduction: Research challenges and recommendations. Health Psychology, 17,
559-568.
Fortin, D. R., & Dholakia, R. R. (2005). Interactivity and vividness effects on social
presence and involvement in a web-based advertisement. Journal of Business
Research, 58, 387-396.
Fox, J., Arena, D., & Bailenson, J. N. (2009). Virtual reality: A survival guide for the
social scientist. Journal of Media Psychology, 21(3), 95-113.
165
Fox, J., & Bailenson, J. N. (2009). Virtual self-modeling: The effects of vicarious
reinforcement and identification on exercise behaviors. Media Psychology, 12, 125.
Fox, J., Bailenson, J. N., & Binney, J. (2009). Virtual experiences, physical behaviors:
The effect of presence on imitation of an eating avatar. PRESENCE:
Teleoperators & Virtual Environments, 18, 294-303.
Fung, J., Richards, C. L., Malouin, F., McFadyen, B. J., & LaMontagne, A. (2006). A
treadmill and motion coupled virtual reality system for gait training post-stroke.
CyberPsychology & Behavior, 9, 157-162.
Garau, M., Slater, M., Pertaub, D.-P., & Razzaque, S. (2005). The responses of people to
virtual humans in an immersive virtual environment. PRESENCE: Teleoperators
& Virtual Environments, 14, 104-116.
Garcia, A. W., & King, A. C. (1991). Predicting long-term adherence to aerobic exercise:
A comparison of two models. Journal of Sport & Exercise Psychology, 13, 394410.
Garcia, K., & Mann, T. (2003). From ‗I wish‘ to ‗I will‘: Social cognitive predictors of
behavioral intentions. Journal of Health Psychology, 8, 347-360.
Geiger, S., & Reeves, B. (1993). We interrupt this program: Attention for television
sequences. Human Communication Research, 19, 368-397.
Gerrard, M., Gibbons, F. X., Lane, D. J., & Stock, M. L. (2005). Smoking cessation:
Social comparison level predicts success for adult smokers. Health Psychology,
24, 623-629.
166
Gilliath, O., McCall, C., Shaver, P. R., & Blascovich, J. (2008). What can virtual reality
teach us about prosocial tendencies in real and virtual environments? Media
Psychology, 11, 259-282.
Ginns, P., Chandler, P., & Sweller, J. (2003). When imagining information is effective.
Contemporary Educational Psychology, 28, 229-251.
Goffman, E. (1959). The presentation of self in everyday life. Garden City, NY:
Doubleday.
Gold, J. I., Belmont, K. A., & Thomas, D. A. (2007). The neurobiology of virtual reality
Pain attenuation. CyberPsychology & Behavior, 10, 536-544.
Goldstein, N. J., & Cialdini, R. B. (2007). The spyglass self: A model of vicarious selfperception. Journal of Personality & Social Psychology, 92, 402-417.
Gollwitzer, P. M., & Oettingen, G. (1998). The emergence and implementation of health
goals. Psychology & Health, 13, 687-715.
Goran, M. I., & Reynolds, K. (2005). Interactive multimedia for promoting physical
activity (IMPACT) in children. Obesity Research, 13, 762-771.
Gratch, J., & Marsella, S. (2005). Lessons from emotion psychology for the design of
Lifelike characters. Applied Artificial Intelligence, 19, 215-233.
Green, M. C. (2006). Narratives and cancer communication. Journal of Communication,
56, S163-S183.
Green, M. C., & Brock, T. C. (2000). The role of transportation in the persuasiveness of
public narratives. Journal of Personality & Social Psychology, 79, 701-721.
167
Greenberg, J., Schimel, J., & Martens, A. (2002) Ageism: Denying the face of the future.
In T. D. Nelson (Ed.), Ageism: Stereotyping and prejudice against older persons
(pp. 27-48). Cambridge, MA: MIT Press.
Gregg, L., & Tarrier, N. (2007). Virtual reality in mental health: A review of the
literature. Social Psychiatry & Psychiatric Epidemiology, 42, 343-354.
Groom, V., Bailenson, J. N., & Nass, C. (2009). The influence of racial embodiment on
Racial bias in immersive virtual environments. Social Influence, 4, 1-18.
Guadagno, R. E., Blascovich, J., Bailenson, J. N., & McCall, C. (2007). Virtual humans
and persuasion: The effects of agency and behavioral realism. Media Psychology,
10, 1-22.
Gutiérrez-Maldonado, J., Ferrer-García, M., Caqueo-Urizar, A., & Letosa-Porta, A.
(2006). Assessment of emotional reactivity produced by exposure to virtual
environments in patients with eating disorders. CyberPsychology & Behavior, 9,
507-513.
Gwaltney, C. J., Metrik, J., Kahler, C. W., & Shiffman, S. (2009). Self-efficacy and
smoking cessation: A meta-analysis. Psychology of Addictive Behaviors, 23, 5666.
Hall, C. R., & Martin, K. A. (1997). Measuring movement imagery abilities: A revision
of the Movement Imagery Questionnaire. Journal of Mental Imagery, 21, 143154.
Hardy, L., & Callow, N. (1999). Efficacy of external and internal visual imagery
perspectives for the enhancement of performance on tasks in which form is
important. Journal of Sport & Exercise Psychology, 21, 95-112.
168
Harris, S. R., Kemmerling, R. L., & North, M. M. (2002). Brief virtual reality therapy for
Public speaking anxiety. CyberPsychology & Behavior, 5, 543-550.
Hayes, A. F., & Krippendorff, K. (2007). Answering the call for a standard reliability
measure for coding data. Communication Methods & Measures, 1, 77-89.
Hausenblas, H. A., Hall, C. R., Rodgers, W. M., & Munroe, K. J. (1999). Exercise
imagery: Its nature and measurement. Journal of Applied Sport Psychology, 11,
171-180.
Hawkins, R. P., Kreuter, M., Resnicow, K., Fishbein, M., & Dijkstra, A. (2008).
Understanding tailoring in communicating about health. Health Education
Research, 23, 454-466.
Hefner, D., Klimmt, C., & Vorderer, P. (2007). Identification with the player character as
determinant of video game enjoyment. In L. Ma, R. Nakatsu, & M. Rauterberg
(Eds.), Lecture notes in computer science (Vol. 4740, pp. 39-48). Berlin:
Springer.
Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological
Review, 94, 319-340.
Higgins, E. T., Roney, C. J. R., Crowe, E., & Hymes, C. (1994). Ideal versus ought
predilections for approach and avoidance: Distinct self-regulatory systems.
Journal of Personality & Social Psychology, 66, 276-286.
Higgins, E. T., Vookles, J., & Tykocinski, O. (1992). Self and health: How ―patterns‖ of
self-beliefs predict types of emotional and physical problems. Social Cognition,
10, 125-150.
169
Hilmert, C. J., Kulik, J. A., & Christenfeld, N. J. S. (2006). Positive and negative opinion
modeling: The influence of another‘s similarity and dissimilarity. Journal of
Personality & Social Psychology, 90, 440-452.
Hoffman, H. G., Patterson, D. R., Seibel, E., Soltani, M., Jewett-Leahy, L., & Sharar, S.
R. (2008). Virtual reality pain control during burn wound debridement in the
hydrotank. Clinical Journal of Pain, 24, 299-304.
Hoffner, C. (1996). Children‘s wishful identification and parasocial interaction with
favorite television characters. Journal of Broadcasting & Electronic Media, 40,
389-403.
Hoffner, C., & Buchanan, M. (2005). Young adults‘ wishful identification with television
characters: The role of perceived similarity and character attributes. Media
Psychology, 7, 325-351.
Horton, D., & Wohl, R. R. (1956). Mass communication and para-social interaction:
Observations on intimacy at a distance. Psychiatry: Journal for the Study of
Interpersonal Processes, 19, 215-229.
Hoyt, C. L., & Blascovich, J. (2007). Leadership efficacy and women leaders‘ responses
To stereotype activation. Group Processes & Intergroup Relations, 10, 595-616.
Hoyt, C. L., Blascovich, J., & Swinth, K. R. (2003). Social inhibition in virtual
environments. PRESENCE: Teleoperators & Virtual Environments, 12, 183-195.
Huang, S.-F., Tsai, P.-Y., Sung, W.-H., Lin, C.-Y., & Chuang, T.-Y. (2008). The
comparisons of heart rate variability and perceived exertion during simulated
cycling with various viewing devices. PRESENCE: Teleoperators & Virtual
Environments, 17, 575-583.
170
Hyde, J., Hankins, M., Deale, A., & Marteau, T. M. (2008). Interventions to increase selfefficacy in the context of addiction behaviours. Journal of Health Psychology, 13,
607-623.
Ijsselsteijn, W. A., de Kort, Y. A. W., Westerink, J., de Jager, M., & Bonants, R. (2006).
Virtual fitness: Stimulating exercise behavior through media technology.
PRESENCE: Teleoperators & Virtual Environments, 15, 688-698.
Ijsselsteijn, W., de Ridder, H., Freeman, J., Avons, S. E., & Bouwhuis, D. (2001). Effects
of stereoscopic presentation, image motion, and screen size on subjective and
objective corroborative measures of presence. PRESENCE: Teleoperators &
Virtual Environments, 10, 298-311.
Ito K. E., Kalyanaraman, S., Brown, J. D., & Miller, W. C. (2008). Factors affecting
avatar use in a STI prevention CD-ROM. Journal of Adolescent Health, 42, S19.
Ivory, J. D., & Kalyanaraman, S. (2007). The effects of technological advancement and
violent content in video games on players‘ feelings of presence, involvement,
physiological arousal, and aggression. Journal of Communication, 57, 532-555.
Jackson, K. M., & Aiken, L. S. (2006). A psychosocial model of sun protection and
sunbathing in young women: The impact of health beliefs, attitudes, norms, and
self-efficacy for sun protection. Health Psychology, 19, 469-478.
Jaret, C., & Reitzes, D. C. (1999). The importance of racial-ethnic identity and social
setting for Blacks, Whites, and multiracials. Sociological Perspectives, 42, 711737.
171
Johnsen, K., Dickerson, R., Rajj, A., Harrison, C., Lok, B., Stevens, A., et al. (2006).
Evolving an immersive medical communication skills trainer. PRESENCE:
Teleoperators & Virtual Environments, 15, 33-46.
Kaiser, H. F. (1974). Little jiffy, mark IV. Educational & Psychological Measurement,
34, 111-117.
Kalichman, S. C., & Coley, B. (1995). Context framing to enhance HIV-antibody-testing
messages targeted to African-American women. Health Psychology, 14, 247-254.
Kalichman, S. C., Kelly, J. A., Hunter, T. L., Murphy, D. A., & Tyler, R. (1993).
Culturally tailored HIV-AIDS risk-reduction messages targeted to AfricanAmerican urban women: Impact on risk sensitization and risk reduction. Journal
of Consulting & Clinical Psychology, 61, 291-295.
Karoly, P. (1993). Mechanisms of self-regulation: A systems view. Annual Review of
Psychology, 44, 23-52.
Kazdin, A. E. (1974). Covert modeling, model similarity, and reduction of avoidance
behavior. Behavior Therapy, 5, 325-340.
Kelman, H. C. (1961). Processes of opinion change. Public Opinion Quarterly, 25, 57-78.
Kennedy, M. G., O‘Leary, A., Beck, V., Pollard, K., & Simpson, P. (2004). Increases in
calls to the CDC National STD and AIDS Hotline following AIDS-related
episodes in a soap opera. Journal of Communication, 54, 287-301.
Kenny, P., Rizzo, A. A., Parsons, T. D., Gratch, J., & Swartout, W. (2007). A virtual
human agent for training novice therapists clinical interviewing skills. Annual
Review of CyberTherapy & Telemedicine, 5, 77-83.
172
King, A. C., Ahn, D. K., Atienza, A. A., & Kraemer, H. C. (2008). Exploring refinements
in targeted behavioral medicine intervention to advance public health. Annals of
Behavioral Medicine, 35, 251-260.
King, A. C., Ahn, D. K., Oliveira, B. M., Atienza, A. A., Castro, C. M., & Gardner, C. D.
(2008). Promoting physical activity through hand-held computer technology.
American Journal of Preventive Medicine, 34, 138-142.
King, A. C., Castro, C., Wilcox, S., Eyler, A. A., Sallis, J. F., & Brownson, R. C. (2000).
Personal and environmental factors associated with physical activity among
different racial-ethnic groups of U.S. middle-aged and older-aged women. Health
Psychology, 19, 354-364.
King, A. C., Friedman, R., Marcus, B., Castro, C., Napolitano, M., Ahn, D., et al. (2007).
Ongoing physical activity advice by humans versus computers: The Community
Health Advice by Telephone (CHAT) trial. Health Psychology, 26, 718-727.
King, A. C., Rejeski, W. J., & Buchner, D. M. (1998). Physical activity interventions
targeting older adults: A critical review and recommendations. American Journal
of Preventive Medicine, 15, 316-333.
King, A. C., Stokols, D., Talen, E., Brassington, G. S., & Killingsworth, R. (2002).
Theoretical approaches to the promotion of physical activity: Forging a
transdisciplinary paradigm. American Journal of Preventive Medicine, 23, S15S25.
Kiousis, S. (2002). Interactivity: A concept explication. New Media & Society, 4, 355383.
173
Klein, W. M. P., & Cerully, J. L. (2007). Health-related risk perception and decisionmaking: Lessons from the study of motives in social psychology. Social &
Personality Psychology Compass, 1, 334-358.
Klimmt, C., Hefner, D., & Vorderer, P. (2009). The video game experience as ―true‖
identification: A theory of enjoyable alterations of players‘ self-perception.
Communication Theory, 19, 351-373.
Konijn, E. A., Bijvank, M. N., & Bushman, B. J. (2007). I wish I were a warrior: The role
of wishful identification in the effects of violent video games on aggression in
adolescent boys. Developmental Psychology, 43, 1038-1044.
Kosslyn, S. M. (1981). The medium and the message in mental imagery: A theory.
Psychological Review, 88, 46-66.
Kosslyn, S. M., Thompson, W. L., & Ganis, G. (2006). The case for mental imagery.
New York: Oxford.
Kreuter, M. W., & Strecher, V. J. (1996). Do tailored behavior change messages enhance
the effectiveness of health risk appraisal? Results from a randomized trial. Health
Education Research, 11, 97-105.
Kreuter, M. W., Strecher, V. J., & Glassman, B. (1999). One size does not fit all: The
case for tailoring print materials. Annals of Behavioral Medicine, 21, 276-283.
Lang, A. (1995). Defining audio/video redundancy from a limited capacity information
processing perspective. Communication Research, 22, 86-115.
Lang, A. (2000). The information processing of mediated messages: A framework for
communication research. Journal of Communication, 50, 46-70.
174
Lang, A., Bolls, P. D., Potter, R. F., & Kawahara, K. (1999). The effects of production
pace and arousing content on the information processing of television messages.
Journal of Broadcasting & Electronic Media, 43, 451-475.
Lang, A., Borse, J., Wise, K., & David, P. (2002). World Wide Web: Orienting to
structural and content features of computer-presented information.
Communication Research, 29, 215-245.
Lang, A., Geiger, S., Strickwerda, M., & Sumner, J. (1993). The effects of related and
unrelated cuts on viewers‘ memory for television: A limited capacity theory of
television viewing. Communication Research, 20, 4-29.
Lang, A., Newhagen, J., & Reeves, B. (1996). Negative video as structure: Emotion,
attention, capacity, and memory. Journal of Broadcasting & Electronic Media,
40, 460-477.
Lang, A., Schwartz, N., Lee, S., & Angelini, J. (2007). Processing radio PSAs:
Production, pacing, arousing content, and age. Journal of Health Communication,
12, 581-599.
Langlois, J. H., Kalakanis, L., Rubenstein, A. J., Larson, A., Hallam, M., & Smoot, M.
(2000). Maxims or myths of beauty? A meta-analytic and theoretical review.
Psychological Bulletin, 126, 390-423.
Lanier, J. (1992). Virtual reality: The promise of the future. Interactive Learning
International, 8, 275-79.
Lanier, J. (2001, April). Virtually there. Scientific American, 284, 66-75.
Lee, K. M. (2004). Presence, explicated. Communication Theory, 14, 27-50.
175
Lee, K. M., Park, N., & Jin, S.-A. (2006). Narrative and interactivity in computer games.
In P. Vorderer & J. Bryant (Eds.), Playing video games: Motives, responses, &
consequences (pp. 259-274). Mahwah, NJ: Erlbaum.
Leiner, D. J., & Quiring, O. (2008). What interactivity means to the user: Essential
insights into and a scale for perceived interactivity. Journal of ComputerMediated Communication, 14, 127-155.
Lewis, M. L., Weber, R., & Bowman, N. D. (2008). ―They may be pixels, but they‘re
MY pixels‖: Developing a metric of character attachment in role-playing video
games. CyberPsychology & Behavior, 11, 515-518.
Li, H., Daugherty, T., & Biocca, F. (2002). Impact of 3-D advertising on product
knowledge, brand attitude, and purchase intention. The mediating role of
presence. Journal of Advertising, 31(3), 43-57.
Lieberman, D. A. (1997). Interactive video games for health promotion: Effects on
knowledge, self-efficacy, social support, and health. In R. L. Street, W. R. Gold,
& T. R. Manning (Eds.), Health promotion and interactive technology (pp. 103120). Mahwah, NJ: Erlbaum.
Linde, J. A., Rothman, A. J., Baldwin, A. S., & Jeffery, R. W. (2006). The impact of selfefficacy on behavior change and weight change among overweight participants in
a weight loss trial. Health Psychology, 25, 282-291.
Liu, Y., & Shrum, L. J. (2003). What is interactivity and is it always such a good thing?
Implications of definition, person, and situation for the influence of interactivity
on advertising effectiveness. Journal of Advertising, 31(4), 53-64.
176
Lombard, M., & Ditton, T. (1997). At the heart of it all: The concept of presence. Journal
Of Computer-Mediated Communication, 3(2), available at
http://jcmc.indiana.edu/vol3/issue2/lombard.html
Loomis, J. M. (1992). Distal attribution and presence. PRESENCE: Teleoperators &
Virtual Environments, 1, 113-119.
Loomis, J. M., Blascovich, J. J., & Beall, A. C. (1999). Immersive virtual environment
technology as a basic research tool in psychology. Behavior Research Methods,
Instruments, & Computers, 31, 557-564.
Maccoby, E. E., & Wilson, W. C. (1957). Identification and observational learning from
films. Journal of Abnormal Social Psychology, 55, 76-87.
Madsen, K. A., Yen, S., Wlasuik, L., Newman, T. B., & Lustig, R. (2007). Feasibility of
a dance videogame to promote weight loss among overweight children and
adolescents. Archives of Pediatric & Adolescent Medicine, 161, 105-107.
Maibach, E., & Flora, J. A. (1993). Symbolic modeling and cognitive rehearsal: Using
video to promote AIDS prevention self-efficacy. Communication Research, 20,
517-545.
Maibach, E., & Murphy, D. A. (1995). Self-efficacy in health promotion research and
practice: Conceptualization and measurement. Health Education Research, 10,
37-50.
Mantovani, F., Castelnuovo, G., Gaggioli, A., & Riva, G. (2003). Virtual reality training
For health-care professionals. CyberPsychology & Behavior, 6, 389-395.
177
Marcus, B. H., Bock, B. C., Pinto, B. M., Forsyth, L. H., Roberts, M. B., Traficante, R.
M. (1998). Efficacy of an individualized, motivationally-tailored physical activity
intervention. Annals of Behavioral Medicine, 20, 174-180.
Marcus, B. H., Owen, N., Forsyth, L. H., Cavill, N. A., & Fridinger, F. (1998). Physical
activity interventions using mass media, print media, and information technology.
American Journal of Preventive Medicine, 15, 362-378.
Markus, H., & Nurius, P. (1986). Possible selves. American Psychologist, 41, 954-969.
Marlatt, G. A., Baer, J. S., & Quigley, L. A. (1995). Self-efficacy and addictive behavior.
In A. Bandura (Ed.), Self-efficacy in changing societies (pp. 289-315). New York:
Cambridge.
Martin, K. A., & Hall, C. R. (1995). Using mental imagery to enhance intrinsic
motivation. Journal of Sport & Exercise Psychology, 17, 54-69.
Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive load in multimedia
learning. Educational Psychologist, 38, 43-52.
McCall, C., Blascovich, J., Young, A., & Persky, S. (2009). Proxemic behaviors as
predictors of aggression against Black (but not White) males in an immersive
virtual environment. Social Influence, 4, 138-154.
McConahay, J. B. (1982). Self-interest versus racial attitudes as correlates of anti-busing
attitudes in Louisville: Is it the buses or the Blacks? Journal of Politics, 44, 692720.
McConahay, J. B., Hardee, B. B., & Batts, V. (1981). Has racism declined in America? It
depends on who is asking and what is asked. Journal of Conflict Resolution, 25,
563-579.
178
McDonald, D. G., & Kim, H. (2001). When I die, I feel small: Electronic game characters
and the social self. Journal of Broadcasting & Electronic Media, 45, 241-258.
Meichenbaum, D. H. (1971). Examination of model characteristics in reducing avoidance
behavior. Journal of Personality & Social Psychology, 17, 298-307.
Meyerowitz, B. E., & Chaiken, S. (1987). The effect of message framing on breast selfexamination attitudes, intentions, and behavior. Journal of Personality & Social
Psychology, 52, 500-510.
Miles, A., Rapoport, L., Wardle, J., Afuape, T., & Duman, M. (2001). Using the mass
media to target obesity: An analysis of the characteristics and reported behaviour
change of participants in the BBC‘s ―Fighting Fat, Fighting Fit‖ campaign. Health
Education Research, 16, 357-372.
Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our
Capacity for processing information. Psychological Review, 63(2), 81-97.
Moore, P. J., Turner, R., Park, C. L., & Adler, N. (1996). The impact of behavior and
Addiction on psychological models of cigarette and alcohol use during pregnancy.
Addictive Behaviors, 21, 645-658.
Morgan, S. E., Movius, L., & Cody, M. J. (2009). The power of narratives: The effect of
entertainment television organ donation storylines on attitudes, knowledge, and
behaviors of donors and nondonors. Journal of Communication, 59, 135-151.
Mraz, R., Hong, J., Quintin, G., Staines, W. R., McIlroy, W. E., Zakzanis, K. K., et al.
(2003). A platform for combining virtual reality experiments with functional
magnetic resonance imaging. CyberPsychology & Behavior, 6, 359-368.
179
Myers, L. B., & Horswill, M. S. (2006). Social cognitive predictors of sun protection
intention and behavior. Behavioral Medicine, 32, 57-63.
Nakamura, L. (2002). Cybertypes: Race, ethnicity, and identity on the Internet. New
York: Routledge.
Napolitano, M. A., & Marcus, B. H. (2002). Targeting and tailoring physical activity
information using print and information technologies. Exercise & Sport Sciences
Reviews, 30, 122-128.
Nigg, C. R. (2003). Technology‘s influence on physical activity and exercise science:
The present and the future. Psychology of Sport & Exercise, 4, 57-65.
Noar, S. M., Benac, C. N., & Harris, M. S. (2007). Does tailoring matter? Meta-analytic
review of tailored print health behavior change interventions. Psychological
Bulletin, 133, 673-693.
Nowak, K. L., & Biocca, F. (2003). The effect of the agency and anthropomorphism on
users‘ sense of telepresence, copresence, and social presence in virtual
environments. PRESENCE: Teleoperators & Virtual Environments, 12, 481-494
Nowak, K. L., Krcmar, M., & Farrar, K. M. (2008). The causes and consequences of
presence: Considering the influence of violent video games on presence and
aggression. PRESENCE: Teleoperators & Virtual Environments, 17, 256-268.
Nowak, K. L., & Rauh, C. (2006). The influence of the avatar on online perceptions of
anthropomorphism, androgyny, credibility, homophily, and attraction. Journal of
Computer-Mediated Communication, 11, 153-178.
180
Nyberg, L., Lundin-Olsson, L., Sondell, B., Backman, A., Holmlund, K., Eriksson, S., et
al. (2006). Using a virtual reality system to study balance and walking in a virtual
outdoor environment: A pilot study. CyberPsychology & Behavior, 9, 388-395.
Okita, S. Y., Bailenson, J. N., & Schwartz, D. L. (2008). The mere belief in social
interaction improves learning. In S. Barab, K. Hay, & D. Hickey (Eds.),
Proceedings of the 8th International Conference for the Learning Sciences.
Mahwah, NJ: Erlbaum.
O‘Leary, A., Jemmott, L. S., & Jemmott, J. B. (2008). Mediation analysis of an effective
sexual risk-reduction intervention for women: The importance of self-efficacy.
Health Psychology, 27, S180-S184.
O‘Toole, R., Playter, R., Krummel, T., Blank, W., Cornelius, N., Roberts, W., et al.
(1998). Assessing skill and learning in surgeons and medical students using a
force feedback surgical simulator. In W. Wells, A. Colchester, & S. Delp (Eds.),
Proceedings of the International Conference for Medical Image Computing &
Computer-Assisted Intervention (MICCAI). Berlin: Springer.
Paffenbarger, R. S., Jr., Wing, A. L., & Hyde, R. T. (1978). Physical activity as an index
of heart attack risk in college alumni. American Journal of Epidemiology, 108,
161-175.
Papa, M. J., Singhal, A., Law, S., Pant, S., Sood, S., Rogers, E. M., et al. (2000).
Entertainment-education and social change: An analysis of parasocial interaction,
social learning, collective efficacy, and paradoxical communication. Journal of
Communication, 50, 31-55.
181
Park, S., & Catrambone, R. (2007). Social facilitation effects of virtual humans. Human
Factors, 49, 1054-1060.
Parsons, T. D., & Rizzo, A. A. (2008). Affective outcomes of virtual reality exposure
therapy for anxiety and specific phobias: A meta-analysis. Journal of Behavior
Therapy & Experimental Psychiatry, 39, 250-261.
Peng, W. (2008). The meditational role of identification in the relationship between
experience mode and self-efficacy: Enactive role-playing versus passive
observation. CyberPsychology & Behavior, 11, 649-652.
Peng, W. (2009). Design and evaluation of a computer game to promote a healthy diet for
young adults. Health Communication, 24, 115-127.
Persky, S., & Blascovich, J. (2007). Immersive virtual environments versus traditional
platforms: Effects of violent and nonviolent video game playing. Media
Psychology, 10, 135-156.
Phinney, J. S. (1992). The multigroup ethnic identity measure: A new scale for use with
diverse groups. Journal of Adolescent Research, 7, 156-176.
Pitts, R. E., Whalen, D. J., O‘Keefe, R., & Murray, V. (1989). Black and White response
to culturally targeted television commercials: A values-based approach.
Psychology & Marketing, 6, 311-328.
Plante, T. G., Aldridge, A., Bogden, R., & Hanelin, C. (2003). Might virtual reality
promote the mood benefits of exercise? Computers in Human Behavior, 19, 495509.
Potter, R. F. (2000). The effects of voice changes on orienting and immediate cognitive
overload in radio listeners. Media Psychology, 2, 147-177.
182
Powell, G. E. (1973). Negative and positive mental practice in motor skill acquisition.
Perceptual & Motor Skills, 37, 312.
Powers, M. B., & Emmelkamp, P. M. G. (2008). Virtual reality exposure therapy for
anxiety disorders: A meta-analysis. Journal of Anxiety Disorders, 22, 561-569.
Price, M., & Anderson, P. (2007). The role of presence in virtual reality exposure
therapy. Journal of Anxiety Disorders, 21, 742-751.
Rains, S. A. (2008). Seeking health information in the information age: The role of
Internet self-efficacy. Western Journal of Communication, 72, 1-18.
Ratan, R. (2010). Self-presence, explicated. Paper to be presented at the 60th Annual
Conference of the International Communication Association, June 22-26,
Singapore.
Ratan, R., Santa Cruz, M., & Vorderer, P. (2008). Multitasking, presence, and selfpresence on the Wii. In L. Moreno (Ed.), Proceedings of the 10th Annual
International Workshop on Presence (pp.167-190). Barcelona, Spain: Starlab
Barcelona S.L.
Reeves, B., Lang, A., Kim, E. Y., & Tatar, D. (1999). The effects of screen size and
message content on attention and arousal. Media Psychology, 1, 49-67.
Reeves, B., & Nass, C. (1996). The media equation: How people treat computers,
television, and new media like real people and places. New York: Cambridge.
Reeves, B., & Read, J. L. (2010). Total engagement: Using games and virtual worlds to
change the way people work and businesses compete. Harvard Business Press.
Reger, G. M., & Gahm, G. A. (2008). Virtual reality exposure therapy for active duty
soldiers. Journal of Clinical Psychology, 64, 940-946.
183
Reihnard, M.-A., Messner, M., & Sporer, S. L. (2006). Explicit persuasive intent and its
impact on success at persuasion: The determining roles of attractiveness and
likeableness. Journal of Consumer Psychology, 16, 249-259.
Renger, R., Steinfelt, V., & Lazarus, S. (2002). Assessing the effectiveness of a
community-based media campaign targeting physical inactivity. Family &
Community Health, 25(3), 18-30.
Rice, R. E. (1992). Task analyzability, use of new media, and effectiveness: A multi-site
exploration of media richness. Organization Science, 3, 475-500.
Rimal, R. N. (2000). Closing the knowledge-behavior gap in health promotion: The
mediating role of self-efficacy. Health Communication, 12, 219-237.
Rimal, R. N. (2001a). Longitudinal influences of knowledge and self-efficacy on exercise
behavior: Tests of a mutual reinforcement model. Journal of Health Psychology,
6, 31-46.
Rimal, R. N. (2001b). Perceived risk and self-efficacy as motivators. Understanding
individuals‘ long-term use of health information. Journal of Communication, 51,
633-654.
Rimal, R. N., Flora, J. A., & Schooler, C. (1999). Achieving improvements in overall
health orientation: Effects of campaign exposure, information seeking, and health
media use. Communication Research, 26, 322-348.
Riva, G. (2002). Virtual reality for health care: The status of research. CyberPsychology
& Behavior, 5, 219-225.
Riva, G. (2005). Virtual reality in psychotherapy: Review. CyberPsychology & Behavior,
8, 220-230.
184
Riva, G., Davide, F., & Ijsselsteijn, W. A. (Eds.) (2003). Being there: Concepts, effects,
and measurement of user presence in synthetic environments. Amsterdam: IOS
Press.
Rogers, E. M., Vaughan, P. W., Swalehe, R. M. A., Rao, N., Svenkerud, P., & Sood, S.
(1999). Effects of an entertainment-education radio soap opera on family planning
behavior in Tanzania. Studies in Family Planning, 30, 193-211.
Rosekrans, M. A. (1967). Imitation in children as a function of perceived similarity to a
social model and vicarious reinforcement. Journal of Personality & Social
Psychology, 7, 307-315.
Rosen, J. C., Silberg, N. T., & Gross, J. (1988). Eating attitudes test and eating disorders
inventory: Norms for adolescent girls and boys. Journal of Consulting & Clinical
Psychology, 56, 305-308.
Rothbaum, B. O., Hodges, L. F., & Kooper, R. (1997). Virtual reality exposure therapy.
Journal of Psychotherapy Practice & Research, 6, 219-226.
Rothbaum, B. O., Hodges, L., Smith, S., Lee, J. H., & Price, L. (2000). A controlled
study of virtual reality exposure therapy for the fear of flying. Journal of
Consulting & Clinical Psychology, 68, 1020-1026.
Rothbaum, B. O., Ruef, A. M., Litz, B. T., Han, H., & Hodges, L. (2003). Virtual reality
exposure therapy of combat-related PTSD: A case study using
psychophysiological indicators of outcome. Journal of Cognitive Psychotherapy,
17, 163-177.
185
Rovniak, L. S., Anderson, E. S., Winett, R. A., & Stephens, R. S. (2002). Social cognitive
determinants of physical activity in young adults: A prospective structural
equation analysis. Annals of Behavioral Medicine, 24(2), 149-156.
Roy, S., Klinger, E., Legeron, P., Lauer, F., Chemin, I., & Nugues, P. (2003). Definition
of a VR-based protocol to treat social phobia. CyberPsychology & Behavior, 6,
411-420.
Rupp, D. E., Vodanovich, S. J., & Credé, M. (2005). The multidimensional nature of
ageism: Construct validity and group differences. Journal of Social Psychology,
145, 335-362.
Sacau, A., Laarni, J., & Hartmann, T. (2008). Influence of individual factors on presence.
Computers in Human Behavior, 24, 2255-2273.
Sallis, J. F., Haskell, W. L., Fortmann, S. P., Vranizan, K. M., Taylor, C. B., & Solomon,
D. S. (1986). Predictors of adoption and maintenance of physical activity in a
community sample. Preventive Medicine, 15, 331-341.
Sallis, J. F., Hovell, M. F., Hofstetter, C. R., Faucher, P., Elder, J. P., Blanchard, J., et al.
(1989). A multivariate study of determinants of vigorous exercise in a community
sample. Preventive Medicine, 18, 20-34.
Sallis, J. F., Pinski, R. B., Grossman, R. M., Patterson, T. L., & Nader, P. R. (1988). The
development of self-efficacy scales for health-related diet and exercise behaviors.
Health Education Research, 3, 283-292.
Scheck, S., Allmendinger, K., & Hamann, K. (2008). The effects of media richness on
multilateral negotiations in a collaborative virtual environment. Journal of Media
Psychology: Theories, Methods, & Applications, 20, 57-66.
186
Schneider, E. F., Lang, A., Shin, M., & Bradley, S. D. (2004). Death with a story: How
story impacts emotional, motivational, and physiological responses to first-person
shooter video games. Human Communication Research, 30, 361-375.
Schooler, C., Chaffee, S. H., Flora, J. A., & Roser, C. (1998). Health campaign channels:
Tradeoffs among reach, specificity, and impact. Human Communication
Research, 24, 410-432.
Schooler, C., Flora, J. A., & Farquhar, J. W. (1993). Moving toward synergy: Media
supplementation in the Stanford Five-City Project. Communication Research, 20,
587-610.
Schooler, C., Sundar, S. S., & Flora, J. A. (1996). Effects of the Stanford Five-City
Project media advocacy program. Health Education Quarterly, 23, 346-364.
Schultheis, M. T., & Rizzo, A. A. (2001). The application of virtual reality technology in
rehabilitation. Rehabilitation Psychology, 46, 296-311.
Schunk, D. H. (1987). Peer models and children‘s behavioral change. Review of
Educational Research, 57, 149-174.
Schunk, D. H., & Hanson, A. R. (1985). Peer models: Influence on children‘s selfefficacy and achievement. Journal of Educational Psychology, 77, 313-322.
Schwarzer, R. (1992). Self-efficacy in the adoption and maintenance of health behaviors:
Theoretical approaches and a new model. In R. Schwarzer (Ed.), Self-efficacy:
Thought control of action (pp. 217-242). Washington, DC: Hemisphere.
Schwarzer, R. (1999). Self-regulatory processes in the adoption and maintenance of
health behaviors. Journal of Health Psychology, 4, 115-127.
187
Schwarzer, R., Luszczynska, A., Ziegelmann, J. P., Scholz, A., & Lippke, S. (2008).
Social-cognitive predictors of physical exercise adherence: Three longitudinal
studies in rehabilitation. Health Psychology, 27, S54-S63.
Schwarzer, R., & Renner, B. (2000). Social-cognitive predictors of health behavior:
Action self-efficacy and coping self-efficacy. Health Psychology, 19, 487-495.
Schwarzer, R., Schüz, B., Ziegelmann, J. P., Lippke, S., Luszczynska, A., & Scholz, U.
(2007). Adoption and maintenance of four health behaviors; Theory-guided
longitudinal studies on dental flossing, seat belt use, dietary behavior, and
physical activity. Annals of Behavioral Medicine, 33(2), 156-166.
Singhal, A., & Rogers, E. M. (1999). Entertainment-education: A communication
strategy for social change. Mahwah, NJ: Erlbaum.
Singhal, A., & Rogers, E. M. (2002). A theoretical agenda for entertainment-education.
Communication Theory, 12(2), 117-135.
Skalski, P., & Tamborini, R. (2007). The role of social presence in interactive agentbased persuasion. Media Psychology, 10, 385-413.
Slater, M. (2004). How colorful was your day? Why questionnaires cannot assess
presence in virtual environments. PRESENCE: Teleoperators & Virtual
Environments, 13, 484-493.
Slater, M., Antley, A., Davison, A., Swapp, D., Guger, C., Barker, C., et al. (2006). A
virtual reprise of the Stanley Milgram obedience experiments, PLoS ONE 1(1):
e39. doi:10.1371/journal.pone.0000039, PLoS ONE.
Slater, M., & Steed, A. (2000). A virtual presence counter. PRESENCE: Teleoperators &
Virtual Environments, 9, 413-434.
188
Sondell, B., Nyberg, L., Eriksson, S., Engström, B., Backman, A., Holmlund, K., et al.
(2005). Altered walking pattern in a virtual environment. PRESENCE:
Teleoperators & Virtual Environments, 14, 191-197.
Spitzer, V. M., & Ackerman, M. J. (2008). The Visible Human® at the University of
Colorado 15 years later. Virtual Reality, 12, 191-200.
Steuer, J. (1992). Defining virtual reality: Dimensions determining telepresence. Journal
of Communication, 42, 73-93.
Stotland, E. (1969). Exploratory investigations of empathy. In L. Berkowitz (Ed.),
Advances in experimental social psychology (Vol. 4, pp. 274-314). New York:
Academic Press.
Strecher, V. J., DeVellis, B. M., Becker, M. H., & Rosenstock, I. M. (1986). The role of
self-efficacy in achieving health behavior change. Health Education Quarterly,
13, 73-92.
Suler, J. R. (2002). Identity management in cyberspace. Journal of Applied
Psychoanalytic Studies, 4, 455-459.
Suler, J. R. (2004). The online disinhibition effect. CyberPsychology & Behavior, 7, 321326.
Sundar, S. S. (2004). Theorizing interactivity‘s effects. The Information Society, 20, 385389.
Sundar, S. S. (2008). Social psychology of interactivity in human-website interaction. In
A. N. Joinson, K. Y. A. McKenna, T. Postmes, & U.-D. Reips (Eds.), The Oxford
handbook of Internet psychology (pp. 89-104). Oxford, UK: Oxford University
Press.
189
Sveistrup, H., McComas, J., Thornton, M., Marshall, S., Finestone H., McCormick, A., et
al. (2003). Experimental studies of virtual reality-delivered compared to
conventional exercise programs for rehabilitation. CyberPsychology & Behavior,
6, 245-249.
Taylor, T. L. (2002). Living digitally: Embodiment in virtual worlds. In R. Schroeder
(Ed.), The social life of avatars: Presence and interaction in shared virtual
environments (pp. 40-62). New York: Springer.
Thalmann, N. M., & Thalmann, D. (1994). Towards virtual humans in medicine: A
prospective view. Computerized Medical Imaging & Graphics, 18, 97-106.
Thayer, R. (1989). The biopsychology of mood and arousal. New York: Oxford.
Timmerman, C. E., & Kruepke, K. A. (2006). Computer-assisted instruction, media
richness, and college student performance. Communication Education, 55, 73104.
Trevino, L. K., Lengel, R. H., & Daft, R. L. (1987). Media symbolism, media richness,
and media choice in organizations: A symbolic interactionist perspective.
Communication Research, 14, 553-574.
Turkle, S. (1995). Life on the screen: Identity in the age of the Internet. New York:
Touchstone.
Valente, T. W., Murphy, S., Huang, G., Gusek, J., Greene, J., & Beck, V. (2007).
Evaluating a minor storyline on ER about teen obesity, hypertension, and 5 a day.
Journal of Health Communication, 12, 551-566.
190
Van Leeuwen, M. L., Veling, H., Van Baaren, R. B., & Dijksterhuis, A. (2009). The
influence of facial attractiveness of imitation. Journal of Experimental Social
Psychology, 45, 1295-1298.
Van Ryn, M., Lytle, L. A., & Kirscht, J. P. (1996). A test of the theory of planned
behavior for two health-related practices. Journal of Applied Social Psychology,
26, 871-883.
Van Schaik, P., Blake, J., Pernet, F., Spears, I., & Fencott, C. (2008). Virtual augmented
exercise gaming for older adults. CyberPsychology & Behavior, 11, 103-106.
Villani, D., Riva, F., & Riva, G. (2007). New technologies for relaxation: The role of
presence. International Journal of Stress Management, 14, 260-274.
Vorderer, P. (2000). Interactive entertainment and beyond. In D. Zillmann & P. Vorderer
(Eds.), Media entertainment: The psychology of its appeal (pp. 21-36). Mahwah,
NJ: Erlbaum.
Vorderer, P., Bryant, J., Pieper, K. M., & Weber, R. (2006). Playing video games as
entertainment. In P. Vorderer & J. Bryant (Eds.), Playing video games: Motives,
responses, & consequences (pp. 1-7). Mahwah, NJ: Erlbaum.
Vorderer, P., Klimmt, C., & Ritterfeld, U. (2004). Enjoyment: At the heart of media
entertainment. Communication Theory, 14, 388-408.
Wallace, L. S., Buckworth, J., Kirby, T. E., & Sherman, W. M. (2000). Characteristics of
exercise behavior among college students: Application of social cognitive theory
to predicting stage of change. Preventive Medicine, 31, 494-505.
Walther, J. B. (1996). Computer-mediated communication: Impersonal, interpersonal,
and hyperpersonal interaction. Communication Research, 23, 3-43.
191
Walther, J. B. (2007). Selective self-presentation in computer-mediated communication:
Hyperpersonal dimensions of technology, language, and cognition. Computers in
Human Behavior, 23, 2538-2557.
Wang, Y., & Morgan, W. P. (1992). The effect of imagery perspectives on the
psychophysiological responses to imagined exercise. Behavioural Brain
Research, 52, 167-174.
Warziski, M. T., Sereika, S. M., Styn, M. A., Music, E., & Burke, L. E. (2008). Changes
in self-efficacy and dietary adherence: The impact on weight loss in the PREFER
study. Journal of Behavioral Medicine, 31, 81-92.
White, A., & Hardy, L. (1995). Use of different imagery perspectives on the learning and
performance of different motor skills. British Journal of Psychology, 86, 169-180.
Winett, R. A., Tate, D. F., Anderson, E. S., Wojcik, J. R., & Winett, S. G. (2005). Longterm weight gain prevention: A theoretically based Internet approach. Preventive
Medicine, 41, 629-641.
Wirth, W., Hartmann, T., Böcking, S., Vorderer, P., Klimmt, C., Schramm, H., et al.
(2007). A process model of the formation of spatial presence experiences. Media
Psychology, 9, 493-525.
Wise, K., & Reeves, B. (2007). The effect of user control on the cognitive and emotional
processing of pictures. Media Psychology, 9, 549-566.
Witmer, B. G., & Singer, M. J. (1998). Measuring presence in virtual environments: A
Presence questionnaire. PRESENCE: Teleoperators & Virtual Environments, 7,
225-240.
192
Witte, K., & Allen, M. (2000). A meta-analysis of fear appeals: Implications for effective
public health campaigns. Health Education & Behavior, 27, 591-615.
Woolfolk, R. L., Parrish, M. W. & Murphy, S. M. (1985). The effects of positive and
negative imagery on motor skill performance. Cognitive Therapy & Research, 9,
335-341.
Wulfert, E., & Wan, C. K. (1995). Safer sex intentions and condom use viewed from a
health belief, reasoned action, and social cognitive perspective. Journal of Sex
Research, 32, 299-311.
Yee, N. (2006). The psychology of massively multi-user online role-playing games:
Motivations, emotional investment, relationships and problematic usage. In R.
Schroeder & A.-S. Axelsson (Eds.), Avatars at work and play: Collaboration and
interaction in shared virtual environments (pp. 187-207). New York: Springer.
Yee, N. & Bailenson, J. N. (2007). The Proteus effect: Self transformations in virtual
reality. Human Communication Research, 33, 271-290.
Yee, N., Bailenson, J. N., & Duchenaut, N. (2009). The Proteus effect: Implications of
transformed digital self-representation on online and offline behavior.
Communication Research, 36, 285-312.
Zanbaka, C., Ulinski, A., Goolkasian, P., & Hodges, L. F. (2007). Responses to virtual
humans: Implications for future interface design. Proceedings of the Conference
on Computer-Human Interaction (CHI) 2007, 1561-1570.
193